UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Blood biomarker panels of the late phase asthmatic response Singh, Amritpal 2016

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
24-ubc_2016_september_singh_amritpal.pdf [ 15.26MB ]
Metadata
JSON: 24-1.0307528.json
JSON-LD: 24-1.0307528-ld.json
RDF/XML (Pretty): 24-1.0307528-rdf.xml
RDF/JSON: 24-1.0307528-rdf.json
Turtle: 24-1.0307528-turtle.txt
N-Triples: 24-1.0307528-rdf-ntriples.txt
Original Record: 24-1.0307528-source.json
Full Text
24-1.0307528-fulltext.txt
Citation
24-1.0307528.ris

Full Text

BLOOD BIOMARKER PANELS OF THE LATE PHASE ASTHMATIC RESPONSE by  Amritpal Singh  B.Sc. Honours, Biology and Mathematics, McMaster University, 2010  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF  DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES (Experimental Medicine)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver)  August 2016  © Amritpal Singh, 2016  ii Abstract Individuals with allergic asthma respond differently, but reproducibly, to allergen inhalation challenge. Some individuals develop an early response only (isolated early responders, ERs) while others also go on to develop a late response (dual responders, DRs). The early asthmatic response is characterized by acute airway contraction immediately following allergen inhalation and resolves within 1-3 hours. 60% of asthmatic individuals go on to develop a late asthmatic response that occurs 3-4 hours after allergen inhalation and is characterized by prolonged airway contraction, cellular infiltration of the airways, chronic inflammation, and airway remodeling. It is not understood why late responses do not develop in all sensitized individuals. In this thesis, I combined the power of the allergen inhalation challenge model to study airway responses with unbiased high throughput technologies and data driven computational methods to delineate molecular differences between ERs and DRs using peripheral blood. I identified many markers that discriminated ERs and DRs such as fibronectin, many pro-inflammatory, anti-inflammatory and cell-specific genes, lipid metabolites and amino acids and the T helper type 17 to T regulatory cell (Th17/Treg) ratio. Transcriptional biomarker panels performed well (AUC ~70%) in predicting at risk/susceptible individuals for the late phase asthmatic response prior to allergen challenge. These panels depicted a heightened pro-inflammatory (activation of nuclear factor (NF)-!B signaling) and dampened anti-inflammatory (reduced expression of decoy and formyl peptide receptors which are involved in host response to pathogens) phenotype in DRs compared to ERs. I developed a statistical method which I used to identify multi-omic biomarker panels (cells, gene transcripts and metabolites). The identified panels achieved a systems view of the underlying molecular interactions highlighting common  iii pathways such as lipid metabolism and valine, leucine and isoleucine metabolism across different biological compartments. I demonstrated that inherent molecular differences in blood exist between ERs and DRs suggesting that some asthmatic individuals display early systemic indications of chronic asthma. These findings may be used to develop blood tests to risk-stratify subjects to improve response to therapies, and may lead to earlier and accurate diagnoses.   iv Preface Each chapter contains material that has either been published in peer-reviewed journals or is currently under preparation for submission. Given the multidisciplinary nature of this thesis some technical terms have been added to a glossary for non-specialists and will appear in a different font.  Chapter 2 contains parts of the following manuscripts (*authors contributed equally): a)! Kam SHY*, Singh A*, He J-Q*, Ruan J, Gauvreau GM, O'Byrne PM, FitzGerald JM, Tebbutt SJ. Peripheral blood gene expression changes during allergen inhalation challenge in atopic asthmatic individuals. Journal of Asthma 2012; 49(3): 219-226.  I analyzed the microarray data, performed the differential expression analysis and generated several plots in the manuscript. I helped write the methods section pertaining to the microarray data analysis. I also helped in addressing comments from reviewers during the peer-review process. Microarray assays were performed by Centre for Translational and Applied Genomics (CTAG) at the BC Cancer Agency (Vancouver, BC, Canada). SHYK, JQH, JR and SJT were responsible for sample preparation. SHYK, JQH and SJT also participated in the data analyses and in the writing of the manuscript. GMG, PMO, JMF and SJT participated in the research design and provision of samples. The Journal of Asthma gives the author the right to re-use original work, in whole or in part in a new publication as the primary author.  b)! Yamamoto M*, Singh A*, Ruan J, Gauvreau GM, O'Byrne PM, Carlsten C, FitzGerald JM, Boulet L-P, Tebbutt SJ. Decreased miR-192 expression in peripheral blood of asthmatic individuals undergoing an allergen inhalation challenge. BMC Genomics 2012; 13(1): 655.  I analyzed the miRNA NanoString data, from differential expression to statistical deconvolution. I generated all figures in the manuscript and helped write the methods and results section. I also  v helped in addressing comments from reviewers during the peer-review process. NanoString assays were performed by NanoString Technologies®. MY, JR and SJT participated in sample preparation and follow-up experiments. MY and SJT also participated in conducting the data analyses. MY, CC and SJT participated in the writing of the manuscript. GMG, PMO, CC, JMF, LPB and SJT participated in the research design and provision of samples. BMC Genomics is an open access journal and published material has been reproduced under the Creative Commons Attribution License 4.0.  c)! Singh A, Cohen Freue GV, Oosthuizen JL, Kam SHY, Ruan J, Takhar M, Gauvreau GM, O'Byrne PM, FitzGerald JM, Boulet L-P, Borchers CH. Tebbutt SJ. Plasma proteomics can discriminate isolated early from dual responses in asthmatic individuals undergoing an allergen inhalation challenge. Proteomics Clinical Applications 2012; 6(9-10): 476-485.  I designed the outline of the manuscript, from discovery to replication. I performed the data analyses, generated all figures and tables and wrote the manuscript. GVCF, JLO, SHYK and MT also conducted data analyses. SHYK and JR performed sample preparation for experiments. GVCF and SJT participated in the writing of the manuscript. GMG, PMO, JMF, LPB and SJT participated in the research design and provision of samples. CHB oversaw the iTRAQ MALDI-TOF/TOF and MRM assays and data preprocessing performed at the University of Victoria-Genome BC Proteomics Centre. As per the Copyright Transfer Agreement, Wiley-Blackwell licenses to the Contributor the right to re-use published material for any publication authored by the Contributor excluding journal articles.  Chapter 3 contains parts of the following manuscripts: 1.! Singh A, Yamamoto M, Kam SHY, Ruan J, Gauvreau GM, O'Byrne PM, FitzGerald JM, Schellenberg R, Boulet L-P, Wojewodka G, Kanagaratham C, De Sanctis J, Radzioch D, Tebbutt SJ. Gene-metabolite expression in blood can discriminate allergen-induced isolated early from dual asthmatic responses. PLOS ONE 2013; 8(7): e67907.  vi  I designed the outline of the manuscript, from discovery to validation. I performed the data analyses, generated all figures and tables and wrote the manuscript. MY, and SHYK also conducted data analyses. SHYK and JR performed sample preparation for experiments. Microarray assays were performed by Centre for Translational and Applied Genomics (CTAG) at the BC Cancer Agency (Vancouver, BC, Canada). Metabolite profiling was performed by Metabolon Inc. (Durham, North Carolina, USA). GW, CK, JS and DR performed the lipid profiling experiments. MY, SHY, RS and SJT participated in the writing and review of the manuscript. GMG, PMO, JMF, LPB and SJT participated in the research design and provision of samples. PLOS ONE is an open access journal and published material has been reproduced under the Creative Commons Attribution (CC BY) license.  2.! Singh A*, Yamamoto M*, Ruan J, Choi JY, Gauvreau GM, Olek S, Hoffmueller U, Carlsten C, FitzGerald JM, Boulet L-P, O'Byrne PM, Tebbutt SJ. Th17/Treg ratio derived using DNA methylation analysis is associated with the late phase asthmatic response. Allergy, Asthma & Clinical Immunology 2014; 10(1): 32.  This publication was featured under the ‘Hot Topics’ section.  I designed the outline of the manuscript, from discovery to replication. I performed the data analyses, generated all figures and tables and wrote the manuscript. MY and JYC also conducted data analyses. JR performed sample preparation for experiments. DNA methylation assays were performed by Epiontis (Berlin, Germany) under the supervision of SO and UH. NanoString Elements assays were performed by NanoString Technologies®. MY, CC and SJT also participated in the writing and review of the manuscript. GMG, PMO, JMF, LPB and SJT participated in the research design and provision of samples. Allergy, Asthma & Clinical  vii Immunology is an open access journal and published material has been reproduced under the Creative Commons Attribution License.  Chapter 4 contains parts of the following manuscript in preparation: Singh A, Shannon CP, Kim YW, Yang CX, Gauvreau GM, FitzGerald JM, Boulet L-P, O’Byrne PM, Tebbutt SJ. Blood biomarker panels of the late phase asthmatic response. Manuscript in preparation.  I designed the outline of the manuscript from discovery to validation. I selected the subject cohorts for both RNA sequencing and NanoString assays. I performed all statistical and computational analyses and wrote the entire manuscript. RNA extraction, RNA quality assessment and sample preparation for samples of the discovery cohort was performed by a previous lab technician JR. RNA sequencing on samples of the discovery cohort was performed by Genome Quebec. RNA extraction, RNA quality assessment and sample preparation for samples of the validation cohort was performed by myself under the supervision of YWK and SJT. CP and CXY helped generate the STAR RNA-Seq dataset. CP helped annotate the Trinity contigs. GMG, PMO, JMF, LPB and SJT participated in the research design and provision of samples.   Chapter 5 contains parts of the following manuscript in preparation: Singh A, Gautier B, Shannon C, Vacher M, Rohart F, Tebbut SJ, Lê Cao K-A. DIABLO – a method for multi-omics integration for biomarker discovery. Manuscript in preparation.  I, along with BG, and K-AL-C developed the method. I wrote the initial scripts for integrative biomarker algorithm as well as the function to run the cross-validation. I analyzed the data, generated the figures and wrote the manuscript. BG participated in writing R functions for the  viii method and cross-validation. CS developed the gene set enrichment analysis pipeline. MV developed and wrote the script for the circus plots which was modified by me. FR re-wrote scripts and implemented them into the mixOmics library. SJT and K-AL-C participated in the reviewing of the manuscript.  For the purposes of this thesis, only pertinent material from published papers was used, and some figures were modified to suite the colour scheme of this thesis. This study was approved by Institutional Review Boards with informed consents obtained in compliance with the Research Ethics Board of each recruiting center (UBCPHC— H09-02114; McMaster—08-225, Laval- REC 20395).  ix Table of Contents Abstract .......................................................................................................................................... ii!Preface ........................................................................................................................................... iv!Table of Contents ......................................................................................................................... ix!List of Tables .............................................................................................................................. xiv!List of Figures ...............................................................................................................................xv!List of Abbreviations and Symbols .......................................................................................... xix!Glossary ...................................................................................................................................... xxi!Acknowledgements .................................................................................................................. xxiv!Dedication ................................................................................................................................. xxvi!Chapter 1: Introduction ................................................................................................................1!1.1! Thesis overview ............................................................................................................. 2!1.2! Asthma ........................................................................................................................... 3!1.2.1! Definition ................................................................................................................ 3!1.2.2! Clinical diagnosis .................................................................................................... 3!1.2.3! Asthma medications ................................................................................................ 4!1.2.4! Phenotypes .............................................................................................................. 5!1.3! Allergen-induced asthmatic responses ........................................................................ 5!1.3.1! History..................................................................................................................... 6!1.3.2! Molecular mechanisms ........................................................................................... 6!1.3.2.1! Allergen sensitization .......................................................................................... 6!1.3.2.2! Early asthmatic response ..................................................................................... 7!1.3.2.3! Late asthmatic response ...................................................................................... 8!1.3.3! Isolated early vs dual asthmatic responders ............................................................ 9!1.3.4! Therapies ............................................................................................................... 10!1.4! High throughput molecular technologies .................................................................. 11!1.4.1! Transcriptomics ..................................................................................................... 11!1.4.2! Proteomics ............................................................................................................. 12!1.4.3! Metabolomics ........................................................................................................ 13!1.5! Asthma biomarkers .................................................................................................... 13!1.6! Hypothesis and specific aims ...................................................................................... 14!1.6.1! Hypothesis............................................................................................................. 14!1.6.2! Specific aims ......................................................................................................... 15!Chapter 2: Establishing a discriminatory signal with respect to allergen inhalation challenge and asthmatic responses .............................................................................................18!2.1! Introduction ................................................................................................................. 18!2.2! Materials and methods ............................................................................................... 19!2.2.1! Subjects ................................................................................................................. 19!2.2.2! Methacholine and allergen inhalation challenge ................................................... 20!2.2.3! Cohorts .................................................................................................................. 21!2.2.4! Blood collection and preparation .......................................................................... 22!2.2.5! Experimental techniques ....................................................................................... 23!2.2.5.1! Transcriptomics ................................................................................................. 23!2.2.5.1.1! Affymetrix microarrays .............................................................................. 23!2.2.5.1.2! NanoString nCounter miRNA assay ........................................................... 23! x 2.2.5.2! Proteomics ......................................................................................................... 24!2.2.5.2.1! isobaric Tag for Relative and Absolute Quantification (iTRAQ) ............... 24!2.2.5.2.2! Multiple reaction monitoring (MRM) ......................................................... 24!2.2.6! Data analysis ......................................................................................................... 25!2.2.6.1! Microarray normalization ................................................................................. 25!2.2.6.2! NanoString data normalization ......................................................................... 25!2.2.6.3! iTRAQ and MRM data preprocessing .............................................................. 26!2.2.7! Statistical methodologies ...................................................................................... 27!2.2.7.1! Differential expression analysis ........................................................................ 27!2.2.7.2! Multivariate methods ........................................................................................ 27!2.2.7.2.1! Multiple linear regression ........................................................................... 27!2.2.7.2.2! Principal Component Analysis ................................................................... 28!2.2.7.2.3! sparse Partial Least Squares Discriminant Analysis (sPLS-DA) ................ 28!2.2.8! Pathway analysis ................................................................................................... 29!2.2.8.1! Ingenuity Pathway Analysis (IPA) ................................................................... 29!2.2.8.2! Gene Go ............................................................................................................ 29!2.3! Results .......................................................................................................................... 29!2.3.1! Changes in mRNA expression in the blood of asthmatics individuals after allergen inhalation challenge .............................................................................................................. 29!2.3.2! Changes in miRNA expression in the blood of asthmatics individuals after allergen inhalation challenge ................................................................................................ 30!2.3.2.1! Differentially expressed miRNA ...................................................................... 30!2.3.2.2! Cell-specific miRNA expression as determined using statistical deconvolution .. 31!2.3.3! Plasma proteomics discriminates dual from isolated early responders ................ 33!2.3.3.1! Reproducibility of iTRAQ ................................................................................ 33!2.3.3.2! Differentially expressed proteins between isolated early responders and dual responders at pre- and post-challenge ............................................................................... 35!2.4! Discussion ..................................................................................................................... 38!2.5! Limitations ................................................................................................................... 42!2.6! Conclusion ................................................................................................................... 43!Chapter 3: Molecular changes in the blood discriminate isolated early from dual asthmatic responders .....................................................................................................................................44!3.1! Introduction ................................................................................................................. 44!3.2! Materials and methods ............................................................................................... 45!3.2.1! Cohort ................................................................................................................... 45!3.2.2! Blood collection and preparation .......................................................................... 47!3.2.3! Experimental techniques ....................................................................................... 47!3.2.3.1! DNA methylation analysis ................................................................................ 47!3.2.3.2! Transcriptomics ................................................................................................. 48!3.2.3.3! Affymetric microarrays ..................................................................................... 48!3.2.3.4! NanoString nCounter Elements assay ............................................................... 48!3.2.3.5! Metabolomics .................................................................................................... 49!3.2.4! Data analysis ......................................................................................................... 49!3.2.4.1! Microarray normalization ................................................................................. 49!3.2.4.2! Metabolomics data preprocessing ..................................................................... 50!3.2.4.3! Lipid profiling ................................................................................................... 50! xi 3.2.5! Statistical methodologies ...................................................................................... 51!3.2.5.1! Multivariate methods ........................................................................................ 51!3.2.5.1.1! Regularized Canonical Correlation Analysis (RCCA) ............................... 51!3.2.5.1.2! Partial Least Squares (PLS) ........................................................................ 51!3.2.5.2! Pathway analysis ............................................................................................... 52!3.3! Results .......................................................................................................................... 52!3.3.1! Differentially expressed genes and metabolites between isolated early and dual responders ............................................................................................................................. 52!3.3.2! Gene-metabolite networks in ERs and DRs in response to allergen challenge .... 57!3.3.3! "-6 and "-3 polyunsaturated fatty acids .............................................................. 60!3.3.4! Th17/Treg ratio is associated with the late phase asthmatic response .................. 61!3.3.5! Association of gene-expression profiles with T helper 17 (Th17) cells, T regulatory (Treg) cells and the Th17/Treg ratio ................................................................... 63!3.4! Discussion ..................................................................................................................... 66!3.5! Limitations ................................................................................................................... 72!3.6! Conclusion ................................................................................................................... 73!Chapter 4: Blood biomarker panels of the late phase asthmatic response .............................74!4.1! Introduction ................................................................................................................. 74!4.2! Materials and methods ............................................................................................... 76!4.2.1! Cohorts .................................................................................................................. 76!4.2.1.1! Discovery cohort ............................................................................................... 76!4.2.1.2! Validation cohort ................................................................................................. 77!4.2.2! Experimental techniques ....................................................................................... 79!4.2.2.1! Blood collection and processing ....................................................................... 79!4.2.2.2! RNA extraction ................................................................................................. 79!4.2.2.3! Transcriptomics ................................................................................................. 80!4.2.2.3.1! RNA Sequencing (RNA-Seq) ..................................................................... 80!4.2.2.3.2! nCounter Elements TagSets ........................................................................ 80!4.2.3! Data analysis ......................................................................................................... 81!4.2.3.1! RNA-Seq data analysis ..................................................................................... 81!4.2.3.2! NanoString data quality control and normalization .......................................... 84!4.2.4! Statistical methodologies ...................................................................................... 84!4.2.4.1! Classification algorithms .................................................................................. 84!4.2.4.2! Deep cross-validation .......................................................................................... 86!4.2.4.3! Classification performance measures ............................................................... 86!4.2.5! Gene-set enrichment analysis ............................................................................... 87!4.3! Results .......................................................................................................................... 88!4.3.1! Changes in cellular frequencies and gene expression between ERs and DRs ...... 88!4.3.1.1! Cell-count changes ............................................................................................ 88!4.3.1.2! Changes in gene expression .............................................................................. 89!4.3.2! Classification performances of pre-challenge biomarker panels .......................... 94!4.3.3! Biomarker candidate selection for screening classifiers ....................................... 97!4.3.4! House-keepers ....................................................................................................... 98!4.3.4.1! NanoString Elements probe design ................................................................... 99!4.3.5! Rediscovery using NanoString Elements platform ............................................. 100!4.3.5.1! Quality control and reproducibility assessment .............................................. 100! xii 4.3.5.2! Optimal biomarker panel size ......................................................................... 104!4.3.6! Validation in an independent cohort ..................................................................... 106!4.3.6.1! UCSC gene-isoforms biomarker panel ........................................................... 108!4.3.6.2! Trinity biomarker panel .................................................................................. 111!4.3.7! Biomarker panels in response to allergen challenge ........................................... 111!4.4! Clinical implications ................................................................................................. 112!4.5! Discussion ................................................................................................................... 113!4.6! Limitations ................................................................................................................. 118!4.7! Conclusion ................................................................................................................. 119!Chapter 5: Novel methods for the integration of multiple omic datasets for biomarker discovery .....................................................................................................................................120!5.1! Introduction ............................................................................................................... 120!5.2! Materials and methods ............................................................................................. 121!5.2.1! Gene expression profiling ................................................................................... 122!5.2.2! Surrogate cell-type frequencies ........................................................................... 122!5.2.3! Metabolomics profiling kit ................................................................................. 123!5.2.4! Pathway enrichment analysis .............................................................................. 124!5.3! Results ........................................................................................................................ 124!5.3.1! Integrative classification method ........................................................................ 124!5.3.2! Pipeline for biomarker discovery ........................................................................ 128!5.3.3! Simulation study ................................................................................................. 129!5.3.4! Study 1: Changes in molecular pathways in response to allergen challenge ...... 133!5.3.5! Study 2: Multi-omic biomarker panel of the late phase asthmatic response ...... 140!5.3.6! Multi-omic biomarker panel in response to allergen inhalation challenge ......... 144!5.4! Discussion ................................................................................................................... 145!5.5! Limitations ................................................................................................................. 149!5.6! Conclusion ................................................................................................................. 150!Chapter 6: Conclusions & future directions ...........................................................................151!6.1! Conclusions ................................................................................................................ 151!6.2! Review of aims ........................................................................................................... 151!6.3! Strengths and limitations ......................................................................................... 153!6.4! Future directions ....................................................................................................... 154!References ...................................................................................................................................155!Appendices ..................................................................................................................................181!Appendix A Supplementary material for Chapter 2 ......................................................... 181!A.1! Determining cell-specific miRNA expression using multiple regression ............... 181!A.2! Partial slopes for both granulocytes and PBMCs in HC and asthmatic individuals (pre- and post-challenge). ................................................................................................... 182!A.3! Optimal number of proteins and components to select for sPLS-DA .................... 183!Appendix B Supplementary material for Chapter 3 ......................................................... 184!B.1! Differentially expressed genes and metabolites at post-challenge (normalized to pre-challenge levels, ratio of post to pre levels). ....................................................................... 184!B.2! Correlation of lymphocyte frequencies obtained using a hematolyzer and DNA methylation analysis. .......................................................................................................... 186!Appendix C Supplementary material for Chapter 4 ......................................................... 187!C.1! Phenotypic classification of flippers ....................................................................... 187! xiii C.2! RNA-Sequencing alignment results ........................................................................ 188!C.3! Establishing lower limit of detection in RNA-Seq data .......................................... 189!C.4! NanoString quality control criteria ......................................................................... 189!C.5! Principal Component Analysis of all RNA-Seq datasets combined ....................... 190!C.6! Overlap between biomarker candidates across all RNA-Seq datasets .................... 191!C.7! NanoString probe sequences ................................................................................... 192!C.8! Trinty contigs mapping to known genes ................................................................. 201!C.9! Representative example of probe design for annotated contigs: comp54057_c0_seq6 (HCLS1) .............................................................................................................................. 202!C.10! Trinty contigs mapping to intronic regions or uncharacterized loci ....................... 203!C.11! Coverage plots and genome alignment of Trintiy contigs that mapped to intronic or uncharacterized loci ............................................................................................................ 204!C.12! 95% attenuation of HBA2 using an inactive probe ................................................ 215!C.13! Equations of the UCSC gene-isoforms and Trinity biomarker panels ................... 216!Appendix D Supplementary material for Chapter 5 ......................................................... 218!D.1! Correlation matrix of cell-types inferred from cell marker genes .......................... 218!D.2! Data Integration Analysis for Biomarker discovery using a Latent variable approach for Omics studies (DIABLO) ................................................................................................. 219!D.3! Determine design matrix ......................................................................................... 220!D.4! Simulation study ..................................................................................................... 221!  xiv List of Tables Table 2.1. Demographics of subjects used for the gene expression analysis ............................... 21!Table 2.2 Demographics of subjects used for the miRNA and protein expression analysis ........ 22!Table 2.3 Differentially expressed proteins identified using iTRAQ at pre-challenge at a BH-FDR of 5% ............................................................................................................................................. 37!Table 3.1 Demographics of subjects used for gene and metabolite profiling ............................... 45!Table 3.2 Demographics of subjects used for lipid profiling ........................................................ 46!Table 4.1 Subject demographics of the discovery cohort ............................................................. 76!Table 4.2 Subject demographics of the validation cohort ............................................................... 77!   xv List of Figures Figure 1.1 Allergen-induced asthmatic responses. ......................................................................... 8!Figure 1.2 Schematic of thesis chapters. ....................................................................................... 16!Figure 2.1 Volcano plots for the two comparisons. ...................................................................... 30!Figure 2.2 MiR-192 expression in whole blood and in peripheral blood mononuclear cells (PBMCs). ...................................................................................................................................... 32!Figure 2.3 Reproducibility of iTRAQ using technical replicates. ................................................ 34!Figure 2.4 Analysis of most discriminatory proteins using sPLS-DA. ......................................... 36!Figure 2.5 Technical replication of fibronectin using LC-MRM/MS. ............................................ 38!Figure 3.1 Lung function measurements during the allergen inhalation challenge. ..................... 53!Figure 3.2 Complete blood counts. ............................................................................................... 54!Figure 3.3 Gene network. .............................................................................................................. 56!Figure 3.4 Network plots highlighting the correlation between #G and #M for isolated early and dual responders. ............................................................................................................................ 59!Figure 3.5 Levels of free docosahexaenoic acid in the plasma of early and dual responders undergoing allergen inhalation challenge. .................................................................................... 61!Figure 3.6 Scatter plots of immune cells quantified using DNA methylation analysis with the corresponding cell-specific gene expression profiles. .................................................................. 62!Figure 3.7 Change in the Th17/Treg ratio in early and dual responders from pre to post-challenge. ...................................................................................................................................... 63!Figure 3.8 Correlation circle depicting the strength of correlation between Treg genes (red) and Th17 genes (blue) with their respective latent variables (Comp 1 and Comp 2). ........................ 65!Figure 3.9 Scatter plots of genes significantly correlated with the Th17/Treg ratio. ................... 66! xvi Figure 4.1 Absolute leukocyte counts. .......................................................................................... 88!Figure 4.2 Relative frequencies of immune cells in the blood. .................................................... 89!Figure 4.3 Transcripts ranked based on BH-FDR for each dataset and time point. ....................... 90!Figure 4.4 Tissue enrichment of differential expressed genes at pre-challenge. .......................... 91!Figure 4.5 Top 20 pathways of the up and down-regulated genes in DRs compared to ERs at pre-challenge. ...................................................................................................................................... 93!Figure 4.6 Workflow of biomarker panel discovery and validation. .............................................. 95!Figure 4.7 Estimate of test performance as measured by the AUC using a 200x5-fold deep cross-validation. ........................................................................................................................................ 96!Figure 4.8 MA plot of biomarker and house-keeping candidates. ................................................ 98!Figure 4.9 Binding densities of samples in the test run of the Elements assay prior to and after attenuation of HBA2. ................................................................................................................... 101!Figure 4.10 Effects of HBA2 saturation on the custom elements assay. ..................................... 102!Figure 4.11 Scatter plot matrix of 87 transcripts measured on aliquots of a single RNA sample across batches and across days. .................................................................................................. 103!Figure 4.12 Histogram of gene-gene correlation of biomarker candidates between NanoString and RNA-Seq corresponding to the different datasets. ............................................................... 104!Figure 4.13 Classification performance of different panel sizes using the NanoString data subsetted according to the different RNA-Seq datasets. ............................................................. 105!Figure 4.14 Classification performance measures in the validation cohort. ................................. 107!Figure 4.15 Comparison of UCSC gene-isoforms and Trinity panel with random classifiers. .. 108!Figure 4.16 Fold-changes of transcripts of validated biomarker panels. .................................... 110! xvii Figure 4.17 Predictive potential of the UCSC gene-isoforms and Trinity biomarker panels 2 hours after challenge. .................................................................................................................. 112!Figure 5.1 Workflow of the DIABLO analysis. ......................................................................... 122!Figure 5.2 Integrative methods for biomarker discovery. ........................................................... 126!Figure 5.3 DIABLO biomarker pipeline. ....................................................................................... 128!Figure 5.4 Covariance matrices varying the strength of correlation between variables. ............ 130!Figure 5.5 Relationship between the design and variables selected in the DIABLO model. ........ 132!Figure 5.6 Error rates of various DIABLO models under different designs, types of variables, strength of correlation, fold-change and noise. ........................................................................... 133!Figure 5.7 Systems approach to molecular changes in blood after allergen inhalation challenge...................................................................................................................................................... 135!Figure 5.8 Regulation of asthma genes in response to allergen inhalation challenge in asthmatics...................................................................................................................................................... 136!Figure 5.9 Regulation of asthma genes in response to allergen inhalation challenge in asthmatics in ERs and DRs. .......................................................................................................................... 139!Figure 5.10 Differential expression of IL-10 between ERs and DRs in response to allergen challenge. .................................................................................................................................... 140!Figure 5.11 Multi-omic biomarker panel predicts ERs and DRs prior to challenge. he circos plot consists of an ideogram consisting of variables color-coded to the different omic datasets. A surrounding line graph depicts the expression of each variable in ERs and DRs, respectively. For a correlation cut-off of 0.5, the links corresponding to positive (red) and negative (blue) associations between variables are plotted. ................................................................................ 142!Figure 5.12 Pathway over-representation analysis using InnateDB. .......................................... 144! xviii Figure 5.13 AUC of the multi-omic biomarker panel post-challenge using a leave-one-out cross-validation. ...................................................................................................................................... 145!  xix List of Abbreviations and Symbols  $ AIS AIC AA AUC  Alpha Allergen induced shift Allergen inhalation challenge Arachidonic acid Area under the receiver operating curve  % bp BH-FDR BALF Beta Base pairs Benjamini Hochberg-False Discovery Rate Bronchoalveolar lavage fluid  ChiP-Seq CIC csSAM Chromatin Immunoprecipitation Clinical Investigator Collaborative Cell-specific Significance Analysis of Microarrays  DIABLO  DHA DR(s) Data Integration Analysis for Biomarker discovery  using a Latent variable approach for Omics studies Docosahexaenoic acid Dual responder(s)  EAR ER(s) Enet & ENCODE EBCs ERCC Early asthmatic response Isolated Early responder(s) Elastic net Epsilon Encyclopedia of DNA Elements Exhaled breath condensates External RNA Control Consortium  FARMS Fc&RI FOV FEV1 FVC FENO Factor Analysis for Robust Microarray Summarization Fc epsilon RI Field of view Forced expiratory volume in 1 second Forced Vital Capacity Fractional nitric oxide  GC-MS Gas chromatography-mass spectrometry  IL IPA Interleukin Ingenuity Pathway Analysis  ! KEGG Kappa Kyoto Encyclopedia of Genes and Genomes   xx ' LAR LTE4 LXA4 Limma LC-MRM/MS LC-MS/MS  Lambda Late asthmatic response Leukotriene E4 Lipoxin A4 Linear Models for Microarray and RNA-Seq data Liquid chromatography mass spectrometric multiple reaction monitoring Liquid chromatography-tandem mass spectrometry NPV Negative Predictive Value  " Omega  PBMCs PC20 PCA PG(s) PGCA PLS p PPV Peripheral Blood Mononuclear Cells Provocative Concentration of methacholine that causes a 20% drop in FEV1 Principal Component Analysis Protein Group(s) Protein Group Code Algorithm Partial Least Squares P-value Positive Predictive Value  qPCR Quantitative real-time polymerase chain reaction  Rf RCCA RIN RMA RNA RSEM RT-qPCR Random forest Regularized Canonical Correlation Analysis RNA Integrity Number Robust MultiArray Average Ribose nucleic acid RNA-Seq by Expectation Maximization Reverse transcription quantitative polymerase chain reaction  sPLS-DA STAR Sparse Partial Least Squares Discriminant Analysis Spliced Transcripts Alignment to a Reference  Th T helper  UBC UCSC University of British Columbia University of California, Santa Cruz   xxi Glossary BENJAMINI HOCHBERG FALSE DISCOVERY RATE (BH-FDR): The BH-FDR is a method to correct for multiple hypothesis testing by controlling the false discovery rate (number of false positives divided by the total number of positive findings). For example, when comparing differences between two groups at a single gene level (or any type of variable), the nominal p-value is the likelihood that this difference occurred by chance alone. When considering a greater number of genes simultaneously for the same comparison, it becomes likely that many will be be different between groups only by chance. The BH-FDR can be considered as an adjusted p-value which takes into account the number of hypothesis tests and the rank of the nominal p-value.  cell-specific Significance Analysis of Microarrays (csSAM): A statistical approach that identifies cell-specific gene expression by determining genes whose association with a particular cell-type (gene expression vs. cell-type frequency) changes by phenotypic group.  Cross-validation: procedure used to estimate the test error of a biomarker panel in the absence of an independent validation cohort.  Data Integration Analysis for Biomarker discovery using a Latent variable approach for Omics studies (DIABLO): A statistical method that identifies correlated subsets of variables from different high-dimensional datasets that best discriminate subjects of different phenotypic groups.   xxii Elastic net (Enet): A predictive model (regularized logistic regression) which assumes a linear relationship between the independent variables (gene expression profiles) and dependent variables (phenotype). A group of correlated variables (e.g. genes) are identified that are highly discriminative with respect to the phenotype.   External RNA Control Consortium (ERCC) spike-in controls: an external set of RNA sequences (not homologous to any known organism) that are added to RNA samples to assess the dynamic range, lower limit of detection of the platform and assess the technical performance of gene expression experiments.  Linear Models for Microarray and RNA-Seq Data (limma): Limma uses linear models to identify differentially expressed genes. Limma uses an empirical Bayes procedure to borrow information across genes to modify gene-wise variances resulting in fewer false positive findings especially in the context of studies with small sample sizes.  Principal Component Analysis (PCA): a statistical method that can be used to reduce the dimensionality of a larger dataset into a smaller number of uncorrelated variables that capture as much variation (information) as possible from the original dataset. These new variables are called principal components and are linear combinations of the original variables. PCA can be used for exploratory data analyses of high-dimensional datasets by visualizing subject and variable similarity. PCA can be used to identify biological (phenotype) and experimental (batch) effects underlying the dataset of interest.   xxiii p-value: probability of finding the observed sample result or more extreme results by chance assuming the null hypothesis is true.   Random forest (Rf): A predictive tree model that detects non-linear patterns in the data to best separate groups of subjects.  Technical replication: Confirmation of findings using the same samples as the original discovery cohort. Replication of fibronectin was performed in Chapter 1.2.3.3.2, using two different proteomic profiling techniques using the same set of individuals.  Statistical deconvolution: A method to estimate cellular frequencies in heterogeneous biological samples. These methods can estimate the relative percentage (between 0 and 1) of cells, or surrogate variables (values are not between 0 and 1) which are more appropriate for comparison between groups of subjects. In both methods, cell marker genes based on pure cellular expression as well as the expression of mixed biological samples are required to estimate cell-specific cell counts or surrogate variables.  Validation: Confirmation of findings using samples independent from the original discovery cohort. Validation of the predictive biomarker panels identified in Chapter 1.4.3.6 was performed using an external independent cohort.  xxiv Acknowledgements I would like to thank my supervisor and mentor Dr. Scott Tebbutt who has supported me throughout my graduate studies. He has allowed me to pursue my interests and provided me with many opportunities to grow and succeed. From tag-teaming a presentation in the land down under, to zipping down mountains on bikes in thunderstorms in Aspen, Colorado, we have travelled the world together on this scientific journey. I would also like to thank my supervisory committee Dr. Gabriela Cohen Freue, Dr. Robert Balshaw and Dr. Robert Schellenberg on their immense support and statistical and clinical feedback. They have gone far beyond the role of committee members, from supervising me on manuscripts, meeting for coffee to discuss mathematical/statistical concepts, and giving me books to study; they were only an email away. I also want to thank Dr. Kim-Anh Le Cao, who has supervised me on a collaborative project during a three month visit to Australia in 2014 and ever since via skype. A huge thank you to our collaborator, Dr. Gail Gauvreau who played a major role in the phenotyping of asthmatic individuals used for this project.   I want to acknowledge the financial support from the Canadian Institutes of Health Research, AllerGen NCE., the BC Lung Association and UBC Faculty of Medicine. I would also like to thank the research coordinators of the AllerGen Clinical Investigator Collaborative for their hard work in recruiting subjects for studies in this thesis. Thank you to all the study participants for making this research possible.  Next, a big thank you to Mr. Casey Shannon, who got me started in programming languages such as R and Linux. The time we spent discussing computational/statistical issues in person, via text or email was time well spent! I also want to thank my colleague Mr. Young Woong Kim for teaching me best practices for experimental laboratory techniques and for being  xxv a vibrant spirit in our lab. Your energy and good vibes are contagious. I also want to thank my previous and current lab members Mr. Jian Ruan, Mr. Masatsugu Yamamoto, and Mrs. Sarah Kam, Ms. Yolanda Yang, Mr. Luka Culbrik, Ms. Amreen Toor.   I would like to extend my warmest thanks to the graduate students, professors and research staff at the Centre for Heart Lung Innovation. I also want to thank Mrs. Jennifer Myers for helping with the R Studio sessions. I also want to thank Mr. Dean English for very helpful IT support and last minute poster printing!  Lastly, I would like to thank my family and friends for their love and support. My Vancouver family Loubna and Kenza and my Ontario family, Channi, Gurpreet, and Mummy. My friends Jaismeet, and Katherine who supported me all the way. Thank you!  xxvi Dedication  To my Mother.    1 Chapter 1:!Introduction Asthma is a chronic inflammatory disorder of the airways that affects 300 million people worldwide yet remains poorly understood due to its complexity and heterogeneity1,2. Asthma complexity arises due to the interaction between the genes and the environment which results in variable asthma phenotypes characterized by airway hyperresponsiveness, airway inflammation and variable airflow obstruction. Asthma heterogeneity has led to further characterization of asthma using both clinical and molecular phenotyping, in order to ultimately tailor the right treatment to the right patient. Clinical phenotypes include broad characterizations such as allergic asthma and exercise-induced asthma, whereas molecular phenotyping enables the characterization of subjects based on molecular profiles, for example, Th2-high versus Th2-low phenotypes3. However, further work is needed to determine the stability of these phenotypes over time and how the underlying biology translates to clinical manifestations4. Studies with well-characterized asthma populations in combination with unbiased high throughput technologies and data-driven approaches may identify discriminatory molecules patterns between phenotypic groups. The use of human and/or animal models that simulate both the inflammatory phenotype and environment exposure may link the underlying pathobiology to the clinical phenotype. This will enable the development of tools that can be used for the diagnosis and monitoring of disease and response to therapy leading to personalized therapeutics. This introduction will provide a review of the current literature with a focus on allergic asthma, the late phase asthmatic response and asthma biomarkers.    2 1.1! Thesis overview The goal of this thesis was to identify biomarker panels that are predictive of the allergen-induced late phase asthmatic response in mild atopic asthmatic individuals. Upon allergen exposure, all individuals with allergic asthma develop an early asthmatic response, but some also go on to develop a late phase asthmatic response. Given that the late response is characteristic of chronic asthma and can be induced in subjects with mild asthma, molecular signatures that are predictive of the late phase response may help screen individuals at risk for chronic asthma, potentially leading to earlier diagnosis and effective therapeutic interventions. To begin, I conducted some pilot studies to determine the effects of the allergen inhalation challenge and asthmatic responses on molecular changes in the blood using high throughput profiling techniques. Given that multiple datasets (e.g. transcriptomics, metabolomics) from the same individuals were being generated, I used data-driven methods to integrate these datasets in order to improve biological insight into the molecular mechanisms of the late phase asthmatic response. Furthermore, I developed and executed a biomarker pipeline from discovery to validation of transcriptomic biomarker panels that were predictive of the late phase response. Lastly, I developed an integrative classification method that combines multiple high-dimensional datasets with various visualizations to aid in the interpretability of this tool. These analyses demonstrate the usefulness of the allergen inhalation challenge model in studying asthmatic responses and provide novel biomarker panels that can help identify asthmatic individuals with increased susceptibility for the late phase asthmatic response.   3 1.2! Asthma 1.2.1! Definition According to the 2015 Global INitiative for Asthma (GINA) report5, “Asthma is a heterogeneous disease, usually characterized by chronic inflammation. It is defined by the history of respiratory symptoms such as wheeze, shortness of breath, chest tightness and cough that vary over time and in intensity, together with variable expiratory airflow limitation.” This heterogeneity is the result of multiple overlapping disease phenotypes such that better phenotyping of patients may lead to improved asthma control and personalized treatments6.  1.2.2! Clinical diagnosis Respiratory symptoms such as wheeze, shortness of breath, cough and chest tightness are indicators of asthma and can be assessed using a physical examination5. The onset of these symptoms is triggered by viral infections, exercise, or environmental exposures (allergen or particulate matter). A detailed medical history as well as family history of asthma or allergies increases the likelihood of asthma diagnosis. Spirometry, a common lung function test is the gold standard to making an asthma diagnosis through the use of the Forced Expiratory Volume in 1 second to Forced Vital Capacity (FEV1/FVC) ratio7. The FEV1/FVC ratio is a predictor of airway obstruction (FEV1/FVC less than 0.75-0.80 in adults and less than 0.90 in children) but can be unreliable if spirometry is not performed according to recommended guidelines. Patients with normal spirometry that present with asthma symptoms can undergo bronchoprovocation testing which measures airway hyperresponsiveness. Methacholine is nebulized and inhaled by the patient using two minutes tidal breathing or five breath dosimeter, in doubling concentrations until a drop in FEV1 of more than 20% results (PC20)2. Concentration values less than 16mg/mL  4 are indicative of significant airway hyperreactivity where the lower the concentration the greater the severity of airway hyperresponsiveness.  1.2.3! Asthma medications Asthma medications include reliever and controller medications that combat the airway hyperresponsiveness and airway inflammation5. Relievers include as-needed inhaled short-acting beta2-agonists (SABA) such as salbutamol which causes smooth muscle relaxation. However, regular use of salbutamol in subjects with mild asthma has been shown to cause airway eosinophilia8. This may be because indications of chronic airway inflammation are present even in patients with intermittent or recent-onset asthma, which require the use of low dose inhaled corticosteroids (ICS). Low dose ICS reduces the underlying airway inflammation, asthma exacerbations, hospitalization and death. A step-wise approach is used to control asthma symptoms of mild, moderate and severe asthmatics. Step 1 consists of treatment with low dose ICS with as-needed use of SABA. Step 2 consists of additional controller options such as leukotriene receptor antagonists (LTRAs). As the severity of the asthma symptoms increases, the dose of ICS is increased as well as the use of long-acting beta2-agonists (LABA). Patients with uncontrolled moderate to severe asthma despite proper adherence to controller medications, may be candidates for add-on treatments such as anti-immunoglobulin IgE (anti-IgE) therapy. Patients can step-up or step-down with respect to their treatment strategy depending on how well they are responding to treatment.   5 1.2.4! Phenotypes Asthma has been characterized into broad phenotypic groups such as allergic asthma, non-allergic asthma, occupational asthma, exercise-induced asthma and many others4–6,2,9. The prevalence of allergic disease in individuals with mild to moderate asthma ranges between 50-95% depending on the ethnicity, location of residence and age of onset10. Sensitization to a particular allergen leads airway asthmatic responses upon subsequent exposures which are characterized by airway contraction, airway hyperresponsiveness and airway inflammation. Non-allergic asthma more likely begins in adulthood instead of childhood, occurs mostly in females, and is more severe9. Occupational asthma is caused by high molecular weight compounds such as animal protein, flour or rubber latex which cause an early or dual asthmatic response or low molecular weight compounds such as plicatic acid which is associated with an isolated late asthmatic response11. Exercise-induced asthma occurs in response to physical activity and resolves within 90 minutes but varies depending on the underlying airway inflammation12. Recent use of unbiased approaches based on statistical cluster analyses of clinical variables related to asthma exacerbations, emergency room visits, hospitalization are yielding additional sub-phenotypes of asthma4.  1.3! Allergen-induced asthmatic responses Sensitization to environmental allergens is a major risk factor for developing asthma 13, especially in children where 88% have tested positive for at least 1 of the following allergens: cat, dog, house dust mite, cockroach, and fungi14. Furthermore, allergen exposure in combination with viral infection (rhinovirus) can lead to asthma exacerbations with associated Th2 inflammation 15,16.  6  1.3.1! History Sixty years ago, Herxheimer identified two distinct components in the airway response to inhaled allergens in allergic asthmatics, naming these components the immediate and late reaction17. Subsequent research suggested that different mechanisms must underlie these two response patterns18,19. It is now firmly established that the immediate reaction, which in this thesis is referred to as the early phase asthmatic response (EAR), is characterized by acute airway contraction, occurs in all sensitized asthmatic individuals. In 50-60% of adults and 80% of children, the early response is followed by a chronic late phase asthmatic response (LAR) characterized by chronic airway contraction, and airway inflammation20–22.  1.3.2! Molecular mechanisms 1.3.2.1! Allergen sensitization Inhaled allergen is taken up by dendritic cells in the airway lumen, airway epithelium or the submucosa and processed into epitopes (short, allergen specific peptide sequences)23. These dendritic cells migrate to the local lymph node or respiratory mucosa and present antigen epitopes to naïve T cells which in the presence of interleukin (IL)-4 or IL-13 (secreted from mast cells, basophils etc.) differentiate into T helper type 2 (Th2) cells. Antigen specific Th2 cells interact with B cells by binding the T Cell Receptor – Cluster of Differentiation 3 (TCR-CD3) complex with the Major Histocompatibility Complex class II (MHC class II) molecules on B cells. CD28 on T cells binds to B7 proteins on B cells causing expression of CD40 ligand (CD40L) on T cells. The CD40L binds the CD40 molecule on B cells and in the presence of IL-4 or IL-13 induces B cell class switching to produce Immunoglobulin E (IgE). IgE binds to the  7 high-affinity IgE receptor (Fc&RI) on mast cells, basophils, dendritic cells, macrophages and monocytes. CD23, a C-type lectin mediates the movement of IgE and antigen-IgE complexes across the epithelium resulting in further activation of Fc&RI on mast cells, basophils, macrophages and dendritic cells24. Furthermore, antigen-IgE complexes bind to CD23 on B cells which present antigens to naïve T cells leading to a process known as epitope spreading23. This mechanism is thought to further propagate the allergic cascade and lead to other allergic conditions.   1.3.2.2! Early asthmatic response Upon allergen exposure in sensitized individuals, the allergen binds to IgE bound to the high-affinity IgE receptors (FcεRI) on mast cells, resulting in the cross-linking of IgE receptors which induces mast cell degranulation25. This causes a release of pre-formed mediators, newly synthesized mediators and de novo transcription of pro-inflammatory mediators. Pre-formed mediators include histamine which increase vascular permeability, and recruit inflammatory mediators to the sites of inflammation. Tryptase is also released which promotes airway remodeling through fibroblast proliferation. Newly synthesized mediators such as cysteinyl leukotrienes (cysLTs) and prostaglandins (e.g. PGD2), induce bronchoconstriction and this is classified as the early asthmatic response (Figure 1.1). In fact, 50% of the drop observed during the EAR can be attributed to cysLTs26. These molecules are also potent chemoattractants of inflammatory cells to the airways. Hours later pro-inflammatory cytokines such as IL-3, IL-4 and IL-5 are released further promoting the infiltration of inflammatory cells into the airways. IL-4 promotes Th2 differentiation as well as the expression of adhesion molecules on endothelial cells. IL-13 induces the release of growth factors to promote airway remodeling and mucus  8 hypersecretion. IL-5 recruits eosinophils to the site of inflammation and promotes their proliferation and survival27. Untreated, the early response usually resolves within 1 to 3 hours following allergen exposure (Figure 1.1)22.  Figure 1.1 Allergen-induced asthmatic responses. In all sensitized individuals, the early response occurs immediately after allergen inhalation resulting in airway contraction (decline in FEV1) and resolves within 1 to 2 hours. 60% of these individuals develop the late phase asthmatic response which peaks at 6-9 hours and is characterized by chronic airway contraction, airway infiltration of inflammatory immune cells, mucus secretion and airway remodeling. Modified from Diamant et al.21 Permission obtained from Rightslink.  1.3.2.3! Late asthmatic response A second bronchoconstriction termed the LAR, occurs 3-4 hours after allergen inhalation in 60% of allergic asthmatic individuals (Figure 1.1). This response is more chronic, and persistent and  9 is characterized by cellular inflammation of the airway, increased bronchovascular permeability, increased mucus secretion, and airway remodelling21,22,25,28. The cellular infiltrate is characterized by the presence of various leukocytes, including eosinophils, mast cells, basophils, lymphocytes, and neutrophils in the airways, resulting in complex interactions leading to the release of a cascade of cytokines and chemokines22,29,30. The late response is partly caused by leukotrienes as shown by the partial attenuation of this response using leukotriene receptor antagonists26. The airway epithelium itself responds to external stimuli through the production of chemokines such as chemokine ligands (CCL2) and CCL20 which recruit monocytes and dendritic cells to the airways31. The airway epithelium also produces cytokines such as thymic stromal lymphopoietin (TSLP), granulocyte-macrophage colony-stimulating factor (GM-CSF), IL-25 and IL-33 which target lung dendritic cells to promote Th2 responses32,33. Thus, the LAR is a product of multiple interacting processes between environmental allergens, epithelial responses, mast cell degranulation and inflammatory responses.  1.3.3! Isolated early vs dual asthmatic responders Allergic asthmatic individuals who exhibit the EAR but not the LAR after allergen inhalation challenge are called isolated early responders (ERs). Those individuals that develop both the EAR and LAR after allergen inhalation are called dual responders (DRs). The allergen inhalation challenge has proven to be a useful model to characterize the physiological and cellular changes between ERs and DRs. This model is used by the AllerGen Clinical Investigator Collaborative1 in order to study the pathobiology of allergen-induced asthmatic responses as well as fast track new therapeutics for allergic asthma (see review by Gauvreau et al.22).                                                 1 http://allergen-nce.ca/research/strategy/cic   10  1.3.4! Therapies The allergen inhalation challenge has proven useful in determining the efficacy of asthma therapies such as inhaled corticosteroids, long acting beta2-agonists, leukotriene receptor antagonists and anti-IgE by measuring outcomes such as airway responses, airway hyperresponsiveness, and airway inflammation21,22. Airway responses include the EAR and LAR, airway hyperresponsiveness which is the drop in lung function associated with methacholine inhalation (causes airway contraction but not an inflammatory response) and airway inflammation which is assessed by measuring inflammatory makers such as sputum eosinophils. Novel therapies have included monoclonal antibodies such as anti-IL-13, anti-IL5, and anti-TSLP, lipid mediators such as 5-lipoxgenase-activation protein inhibitor, leukotriene B4 receptor antagonist and prostaglandin E2 and many others (see Gauvreau et al.22 for a comprehensive review of asthma treatments). Presently, inhaled corticosteroids are the gold standard for asthma therapy, however they may also cause systematic side effects (especially at higher doses) such as growth suppression, reduced bone density, cataracts and glaucoma34. Recently, a nonsteroidal glucocorticoid receptor agonist was developed that attenuated the LAR, reduced airway hyperresponsiveness and sputum eosinophilia35. This work shows promise in the development of novel therapies that possess the anti-inflammatory effects of inhaled corticosteroids with much reduced side effects.   11 1.4! High throughput molecular technologies 1.4.1! Transcriptomics Transcriptomics constitutes the study of coding and non coding ribose nucleic acid (RNA) such as messenger RNA, microRNA, long noncoding RNA etc. Transcriptomic profiling has been used to study the biological pathways implicated in asthma36, identify transcriptional asthma phenotypes37, determine transcriptional changes in response to environmental exposures such as allergen38 and diesel exhaust39, and response to therapy40. Most studies have compared asthmatics with controls or across asthma severity using airway or peripheral blood samples. A study comparing the airway repair mechanisms found slower wound healing compared to controls41. The expression of chloride channel, calcium-activated, family member 1 (CLCA1), periostin (POSTN), and serine peptidase inhibitor B2 (SERPINB2) were the top three up-regulated genes in airway epithelial cells compared to smoker controls42. The study showed that these genes were induced by IL-13 and down-regulated by corticosteroids. Furthermore, subjects with increased expression of these genes (Th2-high phenotype) responded well to inhaled corticosteroids compared to subjects with the Th2-low phenotype3. Comparing gene expression between peripheral blood mononuclear cells (PMBCs) collected during an exacerbation state vs. convalescence showed up-regulation of various inflammatory pathways such as arachidonic acid metabolism, leukocyte migration, innate immunity, adaptive immunity, and complement and coagulation cascade during the exacerbation stage43. A longitudinal study following asthmatic subjects for the course of a year identified distinct set of gene expression signatures during periods of exacerbations compared to convalescence in PBMCs. Significantly activated pathways included the Toll Like Receptor (TLR) signaling pathway, Interferon signaling pathway and modulation of IL-15 during periods of exacerbations44. In wheezing children, CD4+  12 T cells from peripheral blood exhibited up-regulation of stress response, proliferative and apoptosis gene expression compared to healthy controls. These studies indicate that evidence of airway inflammation is present in both pulmonary and peripheral compartments, and that both display complementary immune responses.  1.4.2! Proteomics Proteomic profiling using both unbiased and candidate-based techniques have been applied to airway samples such as induced sputum, exhaled breath condensates (EBCs), bronchoalveolar lavage fluid (BALF) and peripheral blood samples such as serum and plasma2,45. Proteomic profiling of induced sputum in healthy and asthmatic individuals using liquid chromatography-tandem mass spectrometry (LC-MS/MS) implicated the role of defense responses, protease inhibitor activity and complement activation. Furthermore, S100 calcium-binding protein A9 (S100A9), a modulator of mast cell function and eosinophil infiltration46 was up-regulated in induced sputum of subjects with neutrophilic uncontrolled asthma47. Proteomic profiling of BALF after allergen challenge showed a significant up-regulation of many proteins such as cytokines, proteases, complement factors, acute phase proteins and matrix proteins, many of which contribute to asthma pathobiology48. Asthma exacerbations up-regulated alpha-1-antitrypsin and complement component C7 in the plasma of children compared to convalescence49. Plasma fibrinogen and complement C3 fragments significantly changed following allergen inhalation challenge with house dust mite (Dermatophagoides Pteronyssinus)50. These findings indicate that inflammatory changes in the lung may be detected in the periphery and be used to study asthma pathogenesis.   13 1.4.3! Metabolomics Metabolomics is the study of end products of metabolic reactions and show promise in assessing asthma severity, lung function and response to therapy51. Although metabolite profiling has been performed using induced sputum, BALF, blood and urine, it has been increasingly used in EBCs using liquid or gas chromatography mass spectrometry (LC-MS, GC-MS) and nuclear magnetic resonance (NMR)2,52. Urine metabolite profiles enriched in citric acid cycle, and stress on energy metabolism discriminated between children with stable asthma, unstable asthma and controls53. Serum metabolite levels have been shown to correlate with asthma severity where asthmatics with lower FEV1% predicted had higher lipid metabolite levels54. Mice challenged with ovalbumin exhibited significant alterations in metabolite profiles in BALF, many of which were reversed upon Dexamethasome treatment55. Models developed using metabolite levels in EBCs identified strong predictors of asthma, asthma control and inhaled corticosteroid use56. Although metabolomics studies in asthma have been limited, they have shown great promise in improving the understanding of asthma pathobiology. Since metabolic processes are strictly controlled, differences in metabolite levels between ERs and DRs may suggest a loss of homeostasis and an inflammatory imbalance that may be identified through the use of metabolite biomarkers.   1.5! Asthma biomarkers Asthma is a complex and heterogeneous disease such that a single biomarker may not suffice to explain the many different sub-phenotypes of asthma. The multi-allergen IgE test to define atopic asthma is recommended as an asthma biomarker outcome for clinical research for children and adults57. Other biomarker outcomes include total serum IgE, allergen-specific IgE, fractional nitric oxide (FENO), sputum eosinophils, complete blood counts (total eosinophils) and urinary  14 leukotriene E4 (LTE4). FENO is predictive of airway inflammation and is associated with Th2 response58, although the range of FENO values vary greatly in the normal population59. Periostin is an extracellular protein that has recently risen as a systemic biomarker of airway inflammation60 and response to treatment61. The human cartilage glycoprotein (YKL-40) is up-regulated in the serum of asthmatics compared to controls and correlates with asthma severity62. By products of arachidonic acid metabolism make up both pro-inflammatory and anti-inflammatory biomarkers such as leukotriene E4 (LTE4) and lipoxin A4 (LXA4), respectively. Urinary LTE4 are elevated in asthmatics compared to controls63 and further increase during asthma exacerbations64. LXA4 is an anti-inflammatory mediator that is decreased in BALF and peripheral blood granulocytes in asthmatics compared to healthy controls. Single biomarkers, that are currently used to risk-stratify patients to improve response to therapy may benefit from being part of panels with other mediators.  1.6! Hypothesis and specific aims 1.6.1! Hypothesis There are mechanistic differences that distinguish early and dual responders, but it is still unclear what prevents or protects some individuals from dual responses after allergen exposure. I postulate that there are key molecular differences between allergen-induced isolated early responders (ERs) and dual responders (DRs) that are detectable in peripheral blood. I aim to identify these differences prior to (pre) and two hours after (post) allergen inhalation challenge by using unbiased high throughput molecular analysis of blood. Discovering these differences is critical to more accurately and economically discriminate sub-phenotypes of asthmatic individuals, and to develop and translate fit-for-purpose biomarkers of asthmatic phenotypes.  15  1.6.2! Specific aims Chapters 2 to 5 address the following specific aims, as summarized in Figure 1.2:  1)! Determine if the allergen inhalation challenge is a useful model to detect molecular changes in peripheral blood of asthmatic individuals. Chapter 2 begins with pilot work that was conducted in order to assess the utility of the allergen inhalation challenge in identifying molecular differences using transcriptional profiling (messenger RNA and microRNA) of asthmatic individuals pre and post allergen inhalation challenge. Proteomic profiling was also performed in order to identify proteins with significantly different abundances between ERs and DRs.   16  Figure 1.2 Schematic of thesis chapters. Chapter 2 consists of pilot studies which identified significant molecular changes in the blood transcriptome after allergen inhalation challenge. Gene and metabolite profiling was performed in Chapter 3 and molecular changes between ERs and DRs were identified. Biomarker panels were identified using RNA-Seq and validated using the NanoString platform in Chapter 4. Chapter 5 pertains to the development of an integrative classification algorithm which was used to develop a multi-omic biomarker panel predictive of the late phase asthmatic response.  17 2)! Identify molecular differences between isolated early and dual responders prior to and after allergen inhalation challenge. In Chapter 3, gene and metabolite profiling were performed on the same set of individuals in order to identify molecular changes between ERs and DRs pre and post allergen challenge. Gene-metabolite interactions were also identified in ERs and DRs. In addition, the frequencies of various subset of lymphocytes were determined using DNA methylation analysis and compared between ERs and DRs.  3)! Develop and validate biomarker panels using the blood transcriptome that can predict an asthmatic individual’s response to allergen inhalation challenge. Blood biomarker panels using RNA sequencing data were developed and transferred to the NanoString Elements platform. These panels were then validated in an external independent cohort.  4)! Develop an integrative classification method and use it to identify a multi-omic biomarker panel that is predictive of the late phase asthmatic response. A statistical method was developed that enabled the integration of high-dimensional omic datasets, variable selection on each dataset and prediction of multi-group phenotypes. This approach was used to 1) achieve a systems view of molecular interactions in peripheral blood of asthmatics in response to allergen challenge and 2) develop a multi-omic biomarker panel that was predictive of the late phase asthmatic response.   18 Chapter 2:!Establishing a discriminatory signal with respect to allergen inhalation challenge and asthmatic responses 2.1! Introduction High throughput techniques such as transcriptomics (coding and non-coding RNA) and proteomics enable the study of heterogeneous diseases such as asthma. Transcriptomic analysis has been used to investigate the mechanisms of allergen challenge using animal models of asthma65–67. Although allergen inhalation challenge has also been performed using human subjects, these studies have been candidate-based in nature, profiling various immune cells (eosinophils, basophils, dendritic cells, T cells, Tregs and Th17 cells) and related candidate genes (TSLP, IL-13, IL-25, IL-33)22,29,68,69,69–73. A time-series analysis of gene expression in asthmatics during asthma exacerbations identified distinct gene signatures that were altered in peripheral blood mononuclear cells (PBMCs)44. Significantly altered pathways during asthma exacerbations included Toll-like receptor pathway, Interferon response pathway, modulation of IL-15, modulation of B cell Antigen receptor and T cell antigen receptor pathways.  Transcriptomics also includes profiling of non-coding RNA such as miRNA, which regulate various molecular processes through mRNA destabilization and translational repression74. Similar to gene expression studies, miRNA profiling studies using the allergen inhalation challenge model have been carried out mainly using mouse models75,76. These studies highlight the role of miRNAs in various biological processes including apoptosis, inflammation, IL-13 and Th2 mediated responses. Similarly in humans, miRNAs have been shown to modulate allergic responses, smooth muscle hyperresponsiveness and proliferation77,78. The use of proteomics in the field of respiratory diseases such as asthma and chronic obstructive lung disease has been limited and this has been attributed to the poor phenotypic  19 selection of patients79. Studies thus far have compared changes in the proteome between healthy and asthmatic subjects mainly using lung tissue, bronchoalveolar lavage, induced sputum, and exhaled breath condensate, and, less commonly, with plasma and serum80. Since many proteins identified through these studies are involved in multiple biological processes, further studies using unbiased technologies may uncover novel markers of disease etiology in asthma.  Collectively the current literature suggests that the allergen inhalation challenge model is a useful model to study molecular changes of allergic inflammation and that careful subject phenotyping is key to identifying molecular signatures that discriminate between different asthma phenotypes. Therefore, the first aim of this thesis project was to conduct pilot studies to determine whether changes in the whole blood transcriptome and plasma proteome of asthmatic individuals occur in response to allergen inhalation challenge. Furthermore, changes in the proteome between isolated early and dual responders (ERs and DRs) were determined prior to (pre) and two hours after (post) allergen inhalation challenge.  2.2! Materials and methods The Institutional Review Boards of the participating institutions, University of British Columbia (UBC), McMaster University, and Université Laval approved this study (NCT01113697).  2.2.1! Subjects Written informed consents were obtained from participants (18-55 years of age) undergoing allergen inhalation challenges as part of the AllerGen Clinical Investigator Collaborative (CIC). Subjects were non-smokers with mild atopic asthma, free of other lung diseases and any cardiovascular disease and had no viral or respiratory tract infections at least 4 weeks prior to  20 allergen challenge. All subjects had physician diagnosed, clinically stable asthma, with a baseline FEV1 ≥ 70% of predicted value, and baseline methacholine PC20 (provocative concentration of methacholine that causes a 20% drop in FEV1) < 16 mg/mL. All subjects developed the early asthmatic response (at least 20% fall in FEV1 within 2 hours after allergen inhalation). Exclusion criteria included the use of inhaled corticosteroids, and use of other asthma mediation with the exception of infrequently inhaled β2-agonist, which was withheld for 8 hours prior to spirometry measurements (see Diamant et.al.21 for an exhaustive list of inclusion and exclusion criteria).  2.2.2! Methacholine and allergen inhalation challenge Methacholine and allergen challenge were performed on triad visits. One days 1 (pre methacholine) and 3 (post methacholine), methacholine inhalation test was performed in order to determine airway hyperresponsiveness defined as the allergen induced shift [PC20]pre/[PC20]post. Methacholine and allergen inhalation challenge were performed in doubling concentrations using a Wright nebulizer35. Skin prick tests were used to determine allergen sensitivity for each participant. The initial concentration of allergen was determined using a formula81 that incorporates both the skin prick testing and methacholine test outputs. Allergen inhalation challenge was performed on day 2, using allergen extracts in doubling doses until a drop in FEV1 of at least 20% was achieved, subsequently FEV1 was measured at regular intervals up to 7 hours post-challenge. All subjects demonstrated an FEV1 drop of 20% between 0-2 hours (early asthmatic response, EAR) after allergen inhalation challenge. Participants that demonstrated a maximum drop in FEV1 greater than 15% between 3-7 hours after allergen challenge (late asthmatic response, LAR) were classified as dual responders (DRs). Subjects that did not achieve  21 this cut-off but had a drop greater than 10% and an allergen induced shift (Pre methacholine PC20/Post methacholine PC20) greater than or equal to 2 were also classified as DRs.   2.2.3! Cohorts Gene expression analysis was performed using blood samples (pre and post) from nine asthmatic individuals undergoing allergen inhalation challenges (Table 2.1) in order to determine transcriptional changes in the blood in response to allergen inhalation challenge.  Table 2.1. Demographics of subjects used for the gene expression analysis Subject Age (years) Sex Allergen % drop in FEV1 during the early phase (EAR) % drop in FEV1 during the late phase (LAR) Site Blood tube 1 36 M Timothy grass -61.0 -38.0 UBC PAXgene 2 35 F Timothy grass -27.0 -5.0 UBC PAXgene 3 47 M Orchard grass -23.0 -17.0 UBC PAXgene 4 21 F HDMDP -32.1 -12.5 McMaster PAXgene 5 20 F Cat -37.7 -17.3 McMaster EDTA 6 27 F HDMDP -43.3 -16.7 McMaster EDTA 7 23 M Cat -31.4 -15.1 McMaster EDTA 8 60 F Cat -25.5 -6.7 McMaster EDTA 9 22 M HDMDP -22.7 -19.9 McMaster EDTA  Eight (4 isolated early responders and 4 dual responders) additional subjects with mild, atopic asthma recruited at Université Laval were selected for additional pilot studies (Table 2.2). Blood samples from subjects 1-7 (pre and post) were used to determine miRNA changes in the blood due to allergen inhalation challenge. miRNA profiling was also performed for 4 healthy individuals (Mean±SD age of 35.0±10.5, 50% females) recruited at St. Paul’s Hospital. Protein profiling was performed in all eight asthmatic subjects (pre and post).   22 Table 2.2 Demographics of subjects used for the miRNA and protein expression analysis Subject Patient ID Age (years) Sex (M:F) Pre [PC20] (mg/mL) Post [PC20] (mg/L) Allergen-induced shifta % drop in FEV1 EAR LAR  ERs   1 1 28 F 12.8 ND ND -20.3 -4.8 2 2 34 F 2.8 6.1 0.4 -21 -1.5 3 3 27 M 4.5 1.8 2.5 -34.4 0 4 4 42 F 5.3 8.6 0.6 -42.1 -11.1  Mean±SD 32.8±6.9 1:3 6.4±4.4 5.5±3.4 1.2±1.2 29.5±5.3 4.4±2.5  DRs   5 1 23 F 0.3 0.2 1.5 -38.9 -31.8 6 2 26 F 5.1 1.5 3.4 -31.4 -14.9 7 3 49 F 3.6 1.0 3.6 -25.3 -12.6 8 4 26 M 0.9 1.0 0.9 -31.5 -15.6  Mean±SD 31.0±12.1 1:3 2.5±2.3 0.9±0.5 2.4±1.4 31.8±2.8 18.7±4.4 a[PC20]pre/[PC20]post, subjects challenged with cat allergen.  2.2.4! Blood collection and preparation Peripheral blood samples were collected using PAXgene Blood RNA tubes (PreAnalytiX-Qiagen/BD, Valencia, CA, USA) and K2 EDTA Vacutainer® tubes (BD, Franklin Lakes, NJ, USA). Complete blood counts and differentials were obtained using EDTA samples for all study participants. After that, EDTA samples were immediately processed for erythrocytes, buffy coat and plasma. All processed samples and PAXgene tubes were stored at -80°C and sent to the Tebbutt laboratory, Vancouver, Canada. Intracellular RNA was extracted from thawed PAXgene tube samples using the RNeasy Mini Kit according to the manufacturer’s protocols (Qiagen, Chatsworth, CA, USA). Total RNA was isolated from EDTA tube samples following a modified TRIzol-based extraction method. The concentration and quality of RNA were assessed using the NanoDrop 8000 Spectrophotometer (Thermo Scientific, Wilmington, DE, USA) and Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA), respectively.    23 2.2.5! Experimental techniques 2.2.5.1! Transcriptomics 2.2.5.1.1! Affymetrix microarrays Genome-wide expression profiling was performed using Affymetrix Human Gene 1.0 ST (Affymetrix, Santa Clara, CA, USA) arrays by the Centre for Translational and Applied Genomics at the BC Cancer Agency (Vancouver, BC, Canada). All microarray data has been uploaded to the Gene Expression Omnibus (GSE34172).  2.2.5.1.2! NanoString nCounter miRNA assay A total of 734 human and human-associated viral miRNAs were simultaneously assayed in 14 samples (pre and post for subjects 1-7 in Table 2.2) using the nCounter® miRNA Expression Assay Kits (NanoString Technologies, Seattle, WA, USA) at NanoString Technologies. This novel technology is based on multiplexed digital counting of RNA moelcules82 without RNA amplification. Specific miRtag sequences bind to their target miRNA through a ligation reaction via a bridging oligonucleotide (bridge). After removal of the bridge, the miRNA-miRtag complex was hybridized to a color-coded reporter probe (specific to the target miRNA) and a biotinylated capture probe which was used to bind the tag complex to the streptavidin-coated surface of the flow cell. Voltage was applied to stretch the tag complexes along the slide and immobilized using biotinylated anti-5’ oligonucleotides. After immobilization, the voltage was turned off and the image was scanned to obtain counts of miRNAs for each sample.  For each sample, the hybridization reaction consisted of 10µL of the Reporter CodeSet, 10µL of hybridization buffer, 5µL of the Capture ProbeSet and 5µL of total RNA (at a concentration of 20ng/µL) from the sample for a total volume of 30µL. After an 18 hour  24 hybridization period at 65°C, the cartridge (12 samples) was processed using the automatic Prep Station. The cartridge was then scanned using the Digital Analyzer which tabulated the barcode counts for all miRNAs in all samples.  2.2.5.2! Proteomics 2.2.5.2.1! isobaric Tag for Relative and Absolute Quantification (iTRAQ) Prior to tryptic digestion, 14 abundant plasma proteins (Albumin, Immunoglobulins (G, A, M), !1-Antitrpsin, Transferrin, Heptaglobin, !1-Acid Glycoprotein, !2-Macroglobulin, Apolipoprotein A-I, Apolipoprotein A-II, Fibrinogen, Apolipoprotein B, and Complement C3) were depleted from all 16 plasma samples (8 pre and 8 post samples for subjects 1-8 in Table 2.2) using immune-affinity chromatography (Genway Biotech, San Diego, CA). Multiple peptides were labelled (Applied Biosystems, Foster City, CA), pooled and acidified to pH 2.5-3.0 with concentrated phosphoric acid (ACP Chemicals Inc., Montreal, Quebec, Canada). Labelled peptides were separated by 2D liquid chromatography, and spotted directly onto 384-spot MALDI ABI 4800 plates. Peptide-based quantitation was performed using a 4800 MALDI-TOF/TOF mass spectrometer (Applied Biosystems) with the acquisition time ranging from 35 to 40 hours83.  2.2.5.2.2! Multiple reaction monitoring (MRM) Technical replication of candidate proteins identified using iTRAQ was performed using Liquid-chromatography-mass spectrometric multiple reaction monitoring84 (LC-MRM/MS). Plasma samples underwent tryptic digestion after which an acidified stable isotope-labeled peptide mixture with formic acid was added. An AB/MDS Sciex 4000 QTRAP with a nano-electrospray  25 ionization source controlled by Analyst 1.5 software (Applied Biosystems) was used for all LC-MRM/MS assays.  2.2.6! Data analysis 2.2.6.1! Microarray normalization The microarray data was analyzed using the statistical computing program R (version 2.13.0). Raw data was imported and normalized using background correction, quantile normalization and summarization using Factor Analysis for Robust Microarray Summarization85 (FARMS). Irrelevant (non-informative) probe sets were filtered using informative/non-informative (I/NI) calls86. Probe sets were considered “informative” if their corresponding probes reflected the same increase or decrease in mRNA expression across all samples. Therefore only probe sets with consistent expression across probes were considered for downstream analyses. Both FARMS and I/NI calls were performed using the farms R-library (version 1.4.0).  2.2.6.2! NanoString data normalization Total sum normalization of six positive spike-in controls (0.125fM-128fM) was performed to account for differences in assay efficiencies (hybridization, purification, binding etc). For each sample, the mean plus 2 times the standard deviation of 8 negative spike-in controls (used to establish a systematic background threshold) was subtracted from each miRNA count in that sample. Only miRNAs with non-negative counts across all samples were retained, resulting in 163 miRNAs. Total sum normalization was performed such that the counts of all assays were equivalent to the average assay count. Then, miRNAs below 100 counts across all samples (set detection level) were removed, resulting in 72 miRNAs for 18 samples (7 pre, 7 post asthmatic  26 samples and 4 healthy control samples). The data was normalized using a log2 transformation prior to statistical analysis.  2.2.6.3! iTRAQ and MRM data preprocessing iTRAQ: Peptide data was processed using ProteinPilotTM software v3.0 with the integrated ParagonTM search algorithm and ProGroupTM algorithm87 (Applied Biosystems) 3.0.0.0, searching against the International Protein Index (IPI HUMAN v3.69, 87129 entries) database88. All entries in the entire database were queried. The Paragon algorithm was used to estimate parameters for modifications, substitutions, and the number of missed cleavages. ProteinPilot was used to assemble peptide data into protein groups (local) for each iTRAQ experiment containing related redundant proteins, and estimate the protein ratios: levels of subject sample labels relative to the reference sample label. These ratios were computed using the weighted geometric means of individual peptides contributing to the identification of a protein group and were corrected using the Auto Bias option in the ProGroup algorithm. Proteins with an Unused Protein Score greater than 2 (99% confidence) were retained for downstream analyses. The Unused Protein Score indicates the number of unique peptide spectra contributing only to the corresponding identified protein. The protein group code algorithm (PGCA) was used to assemble protein groups (global) by connecting overlapping local protein groups across multiple iTRAQ experiments. PGCA ensured the comparison of related proteins by including homologous proteins, redundant proteins, and proteins from the same families, and if they could not be distinguished based on the observed peptide data. A protein confidence interval of 99% was used in the identification of proteins from the list of observed peptides. MRM: Significant proteins identified using iTRAQ were profiled using MRM using one peptide  27 per protein. MRM data were processed using MultiQuant 1.2 (Applied Biosystems) with the MQL algorithm for peak integration.  2.2.7! Statistical methodologies 2.2.7.1! Differential expression analysis The Linear Models for Microarray and RNA-Seq data89 (limma) R-library was used determine significant changes in gene (mRNA) expression, miRNA expression and protein levels. Limma uses a moderation factor that shrinks the variance90 of each variable (e.g. gene, miRNA, protein) towards a common value in order to obtain a less biased variance estimate for each variable. This approach reduces the number of false positives often observed in studies with small sample sizes. All statistical analyses were implemented using the statistical computing program R (versions 2.13.0 and 2.14.2). The Benjamini-Hochberg False Discovery Rate (BH-FDR) was used to correct for multiple hypothesis testing91.   2.2.7.2! Multivariate methods 2.2.7.2.1! Multiple linear regression Cell-specific miRNA expression was determined using cell-specific analysis of microarrays (csSAM) a multiple linear regression approach92. For each group (pre and post), miRNA expression was regressed onto the relative cell-type frequencies to determine the mean miRNA expression for each cell-type. The miRNA expression level for each cell-type was compared between pre and post time points (see Appendix 6.4A.1 for details).    28 2.2.7.2.2! Principal Component Analysis Principal Component Analysis93,94 (PCA) reduces the dimensionality of a high-dimensional dataset (where the number of variables is much greater than the number of samples, p>>n) by constructing a new set of variables called principal components (linear combinations of the original variables). Principal components are orthogonal (uncorrelated) and only a few are required to capture most of the variability of the original high-dimensional dataset. PCA was performed on centered and scaled data using the mixOmics R library95.  2.2.7.2.3! sparse Partial Least Squares Discriminant Analysis (sPLS-DA) Classification of subjects into their respective phenotypic groups using a subset of variables was performed using sPLS-DA96 from the mixOmics R library. sPLS-DA is an extension of sPLS which identifies a subset of correlated variables from two different datasets X and Y from the same number of samples (n, 16 samples: 8 pre and 8 post) and from p and q number of variables, respectively. In sPLS-DA, the Y matrix is the response vector coded as a n by k dummy matrix (k is the number phenotypic groups). Of note is that the pairing between samples was not specified in this analysis. sPLS-DA is a component based method which also performs variable selection where discriminatory variables have non-zero weights and non-discriminatory variables have zero weights. The number of variables to select per component as well as the number of components can be specified for sPLS-DA by the user. The average misclassification rate using 50x10-fold cross-validation was used to determine the optimal number of components and variables to retain for sPLS-DA.    29 2.2.8! Pathway analysis 2.2.8.1! Ingenuity Pathway Analysis (IPA) Pathway enrichment analysis was performed using Ingenuity Pathway Analysis (IPA-Ingenuity Systems, Redwood City, CA, USA). Differentially expressed genes were uploaded to IPA and the following filters were applied: human immune cells, lung tissue and immune cell lines only.  2.2.8.2! Gene Go Target genes for significant miRNAs were predicted using miRanda and TargetScan miRNA prediction databases. The overlap between the target genes and differentially expressed genes in response to allergen inhalation challenge were used for pathway enrichment using GeneGo MetaCore databases: Functional Enrichment by Ontology and Canonical Modeling.   2.3! Results 2.3.1! Changes in mRNA expression in the blood of asthmatics individuals after allergen inhalation challenge Using a BH-FDR < 5%, 1595 (1126 up- and 469 down-regulated) differential expressed genes were identified using limma comparing pre- and post-challenge samples adjusting for blood collection tube-type, RNA Integrity Number (RIN), site, sex, age, methacholine PC20 at pre-challenge, EAR and LAR. Pathway analysis using IPA identified eukaryotic initiation factor 2 signaling (p = 1.98 x 10-4), interferon signaling (p = 7.83 x 10-4) and phosphatidylinositol 3-kinases/protein kinase B (PI3K/AKT) signaling (p = 1.82 x 10-3) as the top three canonical pathways. The top network identified was characterized by inflammatory response, cellular movement and immune cell trafficking.  30  2.3.2! Changes in miRNA expression in the blood of asthmatics individuals after allergen inhalation challenge 2.3.2.1! Differentially expressed miRNA Two independent linear models were used to determine significant miRNAs for the two comparisons; healthy controls (HC) versus asthmatics (two group comparison) at pre-challenge, and pre- versus post-challenge (paired comparison). MiR-192 was significant in both comparisons at a BH-FDR of 1% (Figure 2.1).   Figure 2.1 Volcano plots for the two comparisons. Left panel: Healthy controls vs. asthmatics at pre-challenge, right panel: pre- vs. post-challenge.  MiR-192 was down-regulated in both comparisons, that is, miR-192 was significantly under-expressed in asthmatics (pre-challenge) compared to HC and decreased following allergen ‐ l og 10p ‐v al ue‐ l og 10p ‐v al uelog2 fold‐change log2 fold‐change 31 inhalation challenge (Figure 2.2A).   2.3.2.2! Cell-specific miRNA expression as determined using statistical deconvolution In order to determine whether miR-192 expression was associated with certain cell-type frequencies, miR-192 expression was regressed onto the cell-type frequencies using multiple linear regression for each group (HC, pre and post) independently. Given the small sample size of HC (n=4), the neutrophil, eosinophil and basophil counts were combined (added) to form a granulocyte group whereas the lymphocyte and monocyte counts were combined into a peripheral blood mononuclear cells (PBMCs) group for each subject. The regression coefficients (partial slopes) representing the mean miR-192 expression for granulocytes and PBMCs were extracted for each group. In order to determine whether the partial slopes for granulocytes and PBMCs (Appendix A.2) were significantly different between groups, a test-statistic for each cell-type between two independent groups was calculated (Appendix A.1). This statistic (similar to a Wald test) compared whether the slopes for two independent groups were different for a given cell-type; that is, for the same increase in frequency of a particular cell-type, was the increase in the mean miRNA expression greater in one group compared to the other? MiR-192 expression at the same frequency of granulocytes was similar between HC and asthmatics (pre- and post-challenge), however, the mean miR-192 expression was significantly (p=0.012) higher in HC than in asthmatics (pre-challenge) for the same frequency of PBMCs independently of the frequency of granulocytes (Figure 2.2B). Although Figure 2.2B shows that miR-192 expression in PBMCs decreases post-challenge which is also seen in whole blood (Figure 2.2A), this change did not reach statistical significance (p>0.05).   32  Figure 2.2 MiR-192 expression in whole blood and in peripheral blood mononuclear cells (PBMCs). A. MiR-192 expression was attenuated in asthmatics compared to HCs and decreased after allergen inhalation challenge. B. MiR-192 expression in PBMCs was significantly lower in asthmatics at baseline compared to HCs.  Based on miRanda and TargetScan miRNA prediction databases, 80 genes (56 up-regulated and 26 down-regulated) from the list of differential expressed genes associated with allergen inhalation challenge (Chapter 2.3.1) were targets of miR-192. Gene set enrichment analysis of the 56 up-regulated genes using MetaCore identified a number of significant (p<0.01) pathways such as the Carbohydrate-Responsive Element Binding Protein (ChREBP) regulation pathway and Interferon (IFN) !/" signaling pathway. Network analysis based on curated biological interactions connected members of the input gene list and identified ontologies involving DNA damage and cell cycle regulation, immunological and stress response.  l og 2mi R‐ 19 2  co de  c ou ntA  BMe an  c el l ul ar   mi R‐ 19 2  ex pr es si onp=0.012HC Pre Post HC Pre Post 33 2.3.3! Plasma proteomics discriminates dual from isolated early responders 2.3.3.1! Reproducibility of iTRAQ Across the 21 samples (8 pre and 8 post as well as 5 technical replicates from dual responder subject 3, DR3, see Table 2.2), 142 protein groups (PGs) were identified. 84/142 PGs were detectable (no missing values) across all samples and were retained for downstream analyses. Six samples of DR3 at post-challenge were used to assess the reproducibility of the iTRAQ technology (two replicates per run). A weak correlation was observed between-run replicates, whereas a much stronger correlation was identified for within-run replicates with the exception of run1 (rep1 and rep2) (Figure 2.3). The median value of PGs across all technical replicates was used for the DR3 post-challenge sample in subsequent analyses.  34  Figure 2.3 Reproducibility of iTRAQ using technical replicates. Pearson correlations between six technical replicates in the iTRAQ dataset (84 proteins x 6 samples). The within-assay correlations were much stronger than the between-assay correlations, with the exception of run1 (DR3.post.rep.1 and DR3.post.rep.2).    35 2.3.3.2! Differentially expressed proteins between isolated early responders and dual responders at pre- and post-challenge No significant (BH-FDR = 5%) proteins were identified comparing pre- and post-challenge samples (paired analysis) for all eight asthmatic individuals.  In a separate linear model, comparing the change in expression in ERs with the change in expression in DRs (#ERs vs. #DRs, where # is post-challenge expression minus pre-challenge expression), no significant proteins were identified (BH-FDR = 5%). However, using sPLS-DA (Appendix 6.4A.3) eight proteins were identified that could classify ERs and DRs (Figure 2.4A). A correlation circle was used to identify the relationship between the proteins selected by sPLS-DA (Figure 2.4B). Briefly, vectors drawn from the origin to each of the points (proteins) allows one to determine the relationship between proteins: 1) if the angle between two vectors is less than 90°, there exists a positive correlation between the two proteins, 2) if the angle between two vectors is greater than 90°, there exists a negative correlation between the two proteins, and 3) if the angle between two vectors is equal to 90°, the correlation between the two proteins is zero. The correlation circle in Figure 2.4B displays the eight PGs into two inversely correlated group (Group1 and Group 2). PGs in each group were positively correlated with PGs in the same group but negatively correlated with PGs in the other group. PGs in Group 1 were all up-regulated in DRs and down-regulated in ERs post allergen inhalation challenge (Figure 2.4C). This is opposite to the pattern observed for PGs in Group 2, where PGs were down-regulated in DRs and up-regulated in ERs after allergen inhalation challenge.    36    Figure 2.4 Analysis of most discriminatory proteins using sPLS-DA. A. Sample clustering using PCA based on the 8 discriminatory proteins identified using sPLS-DA. B. Correlation circle depicting two groups of proteins. C. Proteins in Group 1 and Group 2 showing different fold-changes in ERs and DRs.   Four proteins were significantly different between ERs and DRs prior to allergen inhalation challenge at a BH-FDR of 5% (Table 2.3). Three proteins, alpha-1B-glycoprotein (A1BG), fibronectin (FN1) and transthyretin (TTR) were over expressed in DRs compared to ERs, whereas inter-alpha-inhibitor H4 (ITIH4) was under expressed in DRs compared to ERs.  ●●●●●●●●●●●●●●●●−0.4 −0.2 0.0 0.2 0.4−0.4−0.20.00.20.4PC1 (61.2%)PC2 (38.8%)1234567812345678●●●●DR.PostDR.PreER.PostER.PreFN1KNG1A1BGF12AHSGITIH3SHBGADIPOQ−1.0−0.50.00.51.0−1.0 −0.5 0.0 0.5 1.0Correlation of proteins with Component 1Correlation of proteins with Component 2Correlation Circle PlotsGroup&1Group&2−101ITIH3 ADIPOQ AHSG F12 SHBGProteinFold Change (post minus pre expression)ResponseDRERProteins in Group 1−101A1BG FN1 KNG1ProteinFold Change (post minus pre expression)ResponseDRERProteins in Group 2A. B. C.  37  Table 2.3 Differentially expressed proteins identified using iTRAQ at pre-challenge at a BH-FDR of 5% PG Accession Gene Symbol Protein Name p-value Relative Fold changea 46 IPI:IPI00022895.7 A1BG alpha-1B-glycoprotein  p<0.001 +2.90       16 IPI:IPI00845263.1 IPI:IPI00339225.1 IPI:IPI00339224.2 IPI:IPI00022418.1 IPI:IPI00339223.1 IPI:IPI00339227.4 IPI:IPI00339228.1 IPI:IPI00414283.6 IPI:IPI00479723.4 IPI:IPI00855777.1 IPI:IPI00855785.1 IPI:IPI00867588.1       FN1 fibronectin 1 isoform 2 preproprotein  isoform 5 of Fibronectin  fibronectin 1 isoform 4 preproprotein  isoform 1 of Fibronectin  isoform 3 of Fibronectin  isoform 7 of Fibronectin  isoform 8 of Fibronectin  isoform 9 of Fibronectin  FN1 protein  isoform 14 of Fibronectin  isoform 15 of Fibronectin  isoform 13 of Fibronectin        p<0.001       +2.19     5 IPI:IPI00896413.1 IPI:IPI00218192.3 IPI:IPI00896419.3     ITIH4 inter-alpha (globulin) inhibitor H4 isoform 2 precursor  isoform 2 of Inter-alpha-trypsin inhibitor heavy chain H4c  isoform 1 of Inter-alpha-trypsin inhibitor heavy chain H4      0.001     -0.53 IPI:IPI00922043.2 cDNA FLJ51742, highly similar to Inter-alpha-trypsin inhibitor heavy chain H4  IPI:IPI00944960.1 ITIH4 protein   33 IPI:IPI00855916.1 IPI:IPI00022432.1 IPI:IPI00940791.1  TTR transthyretin  transthyretin  20 kDa protein   0.002  +1.89 aaverage expression in DRs divided by average expression in ERs  These four proteins were further measured using multiple reaction monitoring (MRM), in additional aliquots of the same plasma samples of the discovery cohort. FN1 was the only protein that showed successful technical replication of the discriminatory signal identified using iTRAQ (Figure 2.5).   38  Figure 2.5 Technical replication of fibronectin using LC-MRM/MS. Fibronectin was significantly elevated in DRs compared to ERs using both iTRAQ and MRM.  2.4! Discussion These pilot studies have demonstrated that significant molecular changes can be measured in the blood of asthmatic individuals undergoing allergen inhalation challenge. Inflammatory response, cellular movement, and immune cell trafficking comprised the top network using the list of 1595 differentially expressed genes. This is consistent with the known pathology of allergic asthma, a disease driven by chronic inflammation of the airways. Furthermore, cellular movement and immune cell trafficking are consistent with the characteristics of the late-phase response such as infiltration of immune cells into the airways22.  A limited number of statistically significant candidates were identified using miRNA and proteomic profiling. MiR-192 was significantly lower post allergen inhalation challenge 1.01.52.0FN1ProteinRelative ExpressioniTRAQ0.10.20.30.40.5FN1Proteinpmol/ µ LResponseDRERMRM 39 compared to pre-challenge, as well as in asthmatics compared to healthy control subjects. This suggests that miR-192 levels correlate with the asthma phenotype and are altered in response to allergen inhalation challenge. Another study, looking at the effects of cigarette smoke exposure also showed down-regulation of miR-192 in the lungs of rats97. Among the differentially expressed genes in response to allergen inhalation challenge, predicted target genes of miR-192 were largely up-regulated suggesting the inhibitory role of miR-192 on their expression. These target genes were enriched for functions such as cell cycle and immune response supporting the notion that miRNA can regulate such biological functions in allergic asthma. MiR-192 has been studied in various conditions including cancer and autoimmune diseases. Several reports showed that miR-192 affects cellular proliferation through the p53 pathway, which regulates the cell cycle. The cell cycle checkpoint control genes, p53 and p21 were up-regulated in cells with overexpressed miR-192 in vitro using human cell lines98. Cell cycle regulation in response to DNA damage was one of the top-listed pathways of up-regulated miR-192 gene targets in response to allergen challenge, which may suggest that miR-192 mediates cell cycle regulation of blood cells in response to allergen challenge. In addition, as a biomarker in peripheral blood, miR-192 has been reported to decrease in systemic lupus erythematosus, a systemic autoimmune disease inducing inflammatory responses99. Interestingly, miR-192 expression has reportedly decreased in response to TGF-β and loss of miR-192 correlated with tubulointerstitial fibrosis and reduced renal function in renal biopsies from patients with established diabetic nephropathy100. Allergen inhalation challenge induces up-regulation of TGF-β in the airway epithelium101. TGF-β has been implicated in airway remodeling and inflammation, which are features of chronic asthma. Although the origin of the miRNA needs to be clarified, our data showing down-regulated miR-192 in the blood after allergen inhalation challenge may indicate  40 similar TGF-β derived mechanisms. Since the mechanism of action of miR-192 has not been elucidated in allergic airway diseases, further studies are needed to clarify these mechanisms. Cell specific expression based on statistical deconvolution associated miR-192 with PBMCs, i.e., miR-192 in PBMCs was under-expressed in asthmatic subjects compared to healthy controls (relative frequencies of PBMCs were not different between these groups). miR-192 levels are reportedly different among subsets of lymphocytes, higher in CD4+ T cells and lower in CD8+ T cells, NK and B cells102. Allergen challenge induces a dynamic shift of lymphocyte populations in blood. For example, 24 hours after challenge there was a reduction of peripheral blood CD4+ T lymphocytes from a baseline whereas CD4+ T lymphocytes in bronchoalveolar fluid increased, suggesting lymphocyte recruitment into the respiratory system after allergen challenge103.  Unlike the mRNA and miRNA studies, no proteins were statistically significantly altered in response to allergen inhalation. However, several proteins were found to respond differentially to allergen challenge with respect to the responder group, although none were significant after correction for multiple hypothesis testing. At pre-challenge, alpha-1B-glycoprotein (A1BG), fibronectin (FN1), inter-alpha-inhibitor H4 (ITIH4) and transthyretin (TTR) were differentially expressed comparing ERs and DRs. Previous proteomic studies using human plasma samples have indicated that TTR increases after house dust mite challenge in asthmatic individuals with an immediate phase response (early response) compared to asthmatic non-responders50. On the other hand, ITIH4 was under expressed in the plasma of asthmatic individuals relative to controls104. However, both these proteins were not successful in the technical replication using LC-MRM/MS. It is important to note that ITIH4 was not available as an MRM assay, and therefore a general peptide mapping to inter alpha trypsin inhibitor (ITIH) was used. A1BG, also  41 not validated (p = 0.08), has been previously identified in bronchoalveolar lavage fluid but did not significantly change following segmental allergen challenge48. FN1was the only significant protein in the technical replication using LC-MRM/MS.  FN1 is a glycoprotein that functions in cell adhesion, cell migration, phagocytosis, and as a chemoattractant of inflammatory cells105. The concentration of FN1 is dependent on both age and sex106. Plasma FN1 levels in DRs were double that found in ERs prior to undergoing the allergen inhalation challenge. The protein group corresponding to FN1 remained significant (p = 0.02) in an ordinary least-squares linear model, using age and sex as covariates. Previous studies have shown that airway hyperresponsiveness seen in asthma can be attributed to the cell-mediated contraction of collagen fibrils due to increased FN1 deposition in the extracellular matrix107. In addition, increased deposition of FN1 in the subepithelium leads to the thickening of the airway walls of patients with mild atopic asthma108. Based on these findings, one can speculate that increased FN1 deposition in DRs may help explain the increased airway hyperresponsivesness of these subjects. However, in a previous study using bronchial biopsies from patients with asthma, FN1 expression measured by densitometry was not significantly correlated with lung function measures109. A follow-up study using endobroncial biopsies of atopic mild asthmatic patients and non-atopic healthy controls found that FN1 expression in the airway smooth muscle layer was not different between asthmatics and controls110. Given the differences in the tissues, cohorts and analytical method of FN1 quantification with previous studies, it is difficult to directly compare with the findings in this study.   42 2.5! Limitations Despite the inherent limitation of small sample sizes used for these studies, many significant findings were made. Transcriptional changes (e.g. mRNA and miRNA) may be influenced by changes in the frequencies of various cell-types in blood in response to allergen inhalation. Genome-wide expression profiling of individual cell populations may help alleviate this problem but cell sorting is expensive and the procedure itself may alter gene expression. The statistical deconvolution method used in this chapter attempts to address this constraint by identifying cell-specific expression profiles. Although miR-192 levels in PBMCs were not significantly different between pre and post-challenge, this may be due to the fact that the test statistic which compares the mean miRNA expression for each cell type between two independent groups does not take into account the paired structure of this repeated measures study design. Since it is possible to achieve statistical significance with smaller treatment effects in a paired study design, using an unpaired test statistic may explain why the reduction of miR-192 expression in PBMCs post-challenge compared to pre-challenge was not statistically significant. Accounting for the within individual variation through the use of a mixed-effects model or a multilevel approach with a modification to the test statistic may help improve the statistical significance in repeated measures studies.  The proteins considered in this study were less affected by changes in immune cell frequencies. Although the iTRAQ technology attributes multiple peptides to the same protein, the MRM replication was performed using one peptide per protein. For example, one peptide mapped to FN1, whereas current studies are using up to five peptides in MRM-based multiplexed quantitation of FN1111. Therefore, these findings should be taken with caution given the small size used for discovery.  43  2.6! Conclusion Collectively these pilot studies clearly demonstrate the usefulness of the allergen inhalation challenge model in identifying differential changes in the blood transcriptome and proteome of asthmatic individuals. Using unbiased high throughput omics approaches and through careful phenotyping of asthmatic subjects we have demonstrated that even with limited samples sizes molecular changes between early and dual responders can be identified. Despite the limitations inherent to these exploratory studies, this work warrants further study of asthmatic responses using larger sample sizes.  44 Chapter 3:!Molecular changes in the blood discriminate isolated early from dual asthmatic responders 3.1! Introduction Systems biology approaches combine information from different biological layers in order to help unravel the complexities of disease processes. Biological and molecular processes are comprised of interactions between different biological layers such as the genome, methylome, transcriptome, proteome and metabolome. These interactions are often missed when each omic level is studied in isolation, leading to an increased number of false positives, loss of information and unreliable findings. Recent technological advances coupled with decreasing experimental costs have made it possible to obtain multiple high-dimensional datasets for the same set of individuals in order to study these interactions through systems approaches such as multivariate approaches112,113, Bayesian methods114,115, and network analyses116,117. In the context of asthma, Bayesian networks have been used to associate single nucleotide polymorphisms (SNPs) with asthma and eczema118, predict asthma exacerbations using clinical data119, and predict bronchodilator response using SNPs120. Random forest, a tree based classification algorithm has also been used to predict asthma exacerbations using SNP genotype data, showing an improved performance compared to clinical traits alone121. However, studies combining multiple omics datasets from the same individuals to obtain a holistic view of biological processes in asthma have yet to be carried out122.  This chapter of the thesis focuses on understanding the molecular changes between allergen-induced early and dual asthmatic responses in peripheral blood. Transcriptional and metabolite profiling were performed in order to identify molecular changes in the blood between isolated early and dual responders (ERs and DRs) prior to and after allergen inhalation challenge.  45 Various lymphocyte subsets were quantified using DNA methylation analysis and associated with the late phase response. Data driven approaches were used to integrate datasets to narrow down common underlying molecular processes differentiating ERs and DRs. The findings of this chapter implicate an imbalance between pro- and anti-inflammatory processes that may lead to the development of the late phase asthmatic response.  3.2! Materials and methods 3.2.1! Cohort Details regarding subject inclusion/exclusion criteria, methacholine and allergen inhalation challenges and phenotypic characterization can be found in Chapter 2.2. Fourteen subjects (8 ERs and 6 DRs) made up the discovery cohort (Table 3.1). Subjects 1-4 for both ERs and DRs were also used in Chapter 2 (Table 2.2). Candidate molecules identified in the discovery cohort were assessed in the independent external validation cohort (Table 3.2) of nineteen subjects (8 ERs and 11 DRs).  Table 3.1 Demographics of subjects used for gene and metabolite profiling Patient ID Age (year) Sex (M:F) Pre [PC20] (mg/ mL) Post [PC20] (mg/mL) Allergen Induced Shifta % Fall in FEV1 EAR LAR ERs 1 28 F 12.8 ND ND -20.3 -4.8 2 34 F 2.8 6.1 0.4 -21 -1.5 3 27 M 4.5 1.8 2.5 -34.4 0 4 42 F 5.3 8.6 0.6 -42.1 -11.1 5 29 F 0.4 ND ND -44.3 0 6 31 M 11.8 16 0.7 -24.2 -7.5 7 28 F 9.4 16 0.6 -27.1 -7.1 8 42 M 0.1 ND ND -23 -9 Mean±SD 32.6±6.2 3:5 5.9±4.9 9.7±6.2 1.0±0.9 -29.6±9.5 -5.1±4.2 DRs 1 23 F 0.3 0.2 1.5 -38.9 -31.8 2 26 F 5.1 1.5 3.4 -31.4 -14.9 3 49 F 3.6 1.0 3.6 -25.3 -12.6  46 Patient ID Age (year) Sex (M:F) Pre [PC20] (mg/ mL) Post [PC20] (mg/mL) Allergen Induced Shifta % Fall in FEV1 EAR LAR 4 26 M 0.9 1.0 0.9 -31.5 -15.6 5 27 F 0.6 0.1 6 -48.3 -25.8 6 52 F ND ND ND -33 -27 Mean±SD 33.8±13.0 1:5 2.1±2.1 0.76±0.6 4.7±3.4 -34.7±7.9 -21.3±7.9 p-value 0.84  0.09 0.03 0.19 0.29 0.003 a[PC20]pre/[PC20]post EAR: early asthmatic response, LAR: late asthmatic response ND : not determined, Mean ± SD (standard deviation) Note: all subjects were challenged with cat allergen.  Table 3.2 Demographics of subjects used for lipid profiling  Age (year) Sex M:F Allergen Pre [PC20] (mg/ mL) Post [PC20] (mg/mL) Allergen Induced Shift % Fall in FEV1 Early Late ERs 1 25 M Timothy Grass 0.08 ND ND -25.24 -11 2 36 F Timothy Grass 0.64 ND ND -19.66 4.41 3 21 F Ragweed 6.96 2.93 0.421 -33.1 2.11 4 33 M Grass 13.3 21.1 1.587 -50 -2.5 5 21 F Cat 0.59 ND ND -39.66 -1.72 6 30 F Grass 1.81 ND ND -39.8 -11.7 7 21 F Cat 9.6 ND ND -44.96 -12.4 8 43 M Cat 1.94 ND ND -25.6 -10.8 Mean ± SD 28.8±8.2 3:5  4.4±5.0 12±12.9 1.0±0.8 -34.8±10.6 -5.5±6.8 DRs 9 56 F Cat 0.16 0.3 1.875 -23.22 -21.3 10 37 M Timothy Grass 3.26 0.08 0.025 -60.61 -36.9 11 48 M Orchard Grass 0.28 0.79 2.821 -25.16 -19.4 12 44 M Cat hair 1.45 0.36 0.248 -41.33 -21.3 13 40 M Grass 6.9 2.38 0.345 -21.21 -17.6 14 19 M Grass 15.56 2.36 0.152 -42.99 -26.2 15 21 M Ragweed 1.5 1.15 0.767 -55.73 -16.3 16 21 F Cat 0.42 0.23 0.548 -44.06 -31.6 17 43 F Cat 0.28 0.18 0.643 -26.92 -17.3 18 21 F Cat 5.25 3.78 0.72 -47.22 -11.5 19 20 F Grass 5.34 1.64 0.307 -23.61 -16.7  47  Age (year) Sex M:F Allergen Pre [PC20] (mg/ mL) Post [PC20] (mg/mL) Allergen Induced Shift % Fall in FEV1 Early Late Mean ± SD 33.6±14 6:5  3.7±4.6 1.2±1.2 0.8±0.8 -37.5±14 -21.5±7.4 p-value 0.34   0.76 0.44 0.76 0.64 0.0002   3.2.2! Blood collection and preparation Blood samples were obtained prior to and two hours after allergen inhalation challenge and processed as previously described in Chapter 2.2.4. Total DNA was isolated from whole blood or buffy coat from EDTA tubes using QIAamp DNA Blood Mini Kit (Qiagen) according to the manufacturer’s protocol.  3.2.3! Experimental techniques 3.2.3.1! DNA methylation analysis Cell counting of lymphocyte subsets was performed by Epiontis (Berlin, Germany) using quantitative real-time polymerase chain reaction (qPCR) based DNA methylation analysis123,124. Briefly, bisulphite conversion of genomic DNA was performed, which resulted in either 5’-Cytosine-phospate-Guanine-3’ (CpG)-variants (if DNA was methylated) or TpG-variants (unmethylated CpG-variants)125. Each qPCR assay was specific for either the unmethylated FOXP3 Treg specific demethylated region (TSDR) for Tregs, the unmethylated CD3D/G for T cells, the unmethylated IL17A for Th17 cells or the B cell specific unmethylated gene region (propriety Epiontis data) for B cells. Another qPCR assay was specific for a region in the GAPDH control gene, a target that is unmethylated in all cells. The GAPDH qPCR assay serves as a “load control” as it estimates the number of “total cells” in a given sample. The percentage of Treg, T, B and Th17 cells (or the proportion of gene-specific DNA) was determined as the  48 ratio of cell specific TpG-copies and GAPDH TpG-copies. For Tregs only, the ratio was multiplied by 2 in order to correct for the fact that each cell has two copies of the unmethylated GAPDH gene but each Treg has only one copy of the unmethylated FOXP3 gene.   3.2.3.2! Transcriptomics 3.2.3.3! Affymetric microarrays Genome-wide expression profiling was performed using Affymetrix Human Gene 1.0 ST (Affymetrix, Santa Clara, CA, USA) arrays by the Centre for Translational and Applied Genomics (CTAG) at the BC Cancer Agency (Vancouver, BC, Canada). All microarray data was uploaded to the Gene Expression Omnibus (GSE40240).  3.2.3.4! NanoString nCounter Elements assay Technical replication of selected genes was performed using a new digital technology, nCounter Elements (NanoString, Seattle, USA) by NanoString Technologies. nCounter Elements allows users to combine nCounter Elements General Purpose Reagents (GPRs) with unlabeled probes that target specific genes of interest2. 100 ng of each RNA sample was added to the TagSet in hybridization buffer and incubated at 65°C for 16 hours. The TagSet consisted of a reporter probe and capture probe that hybridize to the user designed gene-specific probe A and probe B complex. Automated processing per cartridge on the Prep Station (high sensitivity protocol) occurred for 3 hours. After a 2.5 hour scan per cartridge, counts were acquired from the GEN2 Digital Analyzer.                                                  2 www.nanostring.com/elements   49 3.2.3.5! Metabolomics Candidate metabolite profiling was performed by Metabolon Inc. (Durham, North Carolina, USA). Samples were extracted and prepared for mass spectrometry analysis using Metabolon’s standard solvent extraction method. The extracted samples were divided into equal parts for gas chromatography mass spectrometry (GC-MS) and liquid chromatography tandem mass spectrometry (LC-MS/MS).  3.2.4! Data analysis 3.2.4.1! Microarray normalization For sections 3.3.1 and 3.3.2 preprocessing of microarray data included the Robust MultiArray Average (RMA) background correction, quantile normalization and summarization using FARMS (see Chapter 2.2.6.1). I/NI calls were used to filter out non-informative probe sets (inconsistent expression of samples across probes). For section 3.3.5 the microarray data was normalized using RMA. The difference is normalization is mainly with respect to the summarization and filtering step. I/NI calls filtered out the majority of the probe sets, whereas RMA retained all probe sets for downstream analysis. The former dataset was used to narrow down candidates for further validation therefore a more stringent filtering threshold was used. The latter dataset was used to identify cell-specific genes for the purposes of exploratory analyses, therefore all genes were used. Microarray data normalization was performed using the statistical computing program R (version 2.14.2).    50 3.2.4.2! Metabolomics data preprocessing Preprocessing of mass spectrometry data including data extraction, peak-identification and data preprocessing for quality control and compound identification was performed by Metabolon Inc. (Durham, North Carolina, USA). The hardware and software foundations for these informatics components were the LAN backbone, and a database server running Oracle 10.2.0.1 Enterprise Edition.  3.2.4.3! Lipid profiling Arachidonic acid and docosahexaenoic acid were profiled using plasma samples in a validation cohort of 19 study participants by collaborators at McGill University, Montreal, Canada. Briefly, plasma was first suspended in 1 mM butylated hydroxyanisole in chloroform and methanol (2:1) in order to protect the integrity of the samples. Following a previously described method126, lipids were extracted from these plasma samples using 1.9 ml of chloroform/methanol (2:1 v/v) and 1 ml of cold water. From the organic phase, aliquots were used for phospholipid analysis by thin-layer chromatography127. Lipids were separated from samples by thin-layer chromatography and detected by iodine. After scraping from the plate, the fatty acids were esterified using diazomethane and the esters were quantified using gas chromatography/mass spectrometry (Hewlett Packard5880A, WCOT capillary column (Supelco-10, 35 m60.5 mm, 1 mm thick)) using commercial standards (Sigma-Aldrich, Oakville, ON, Canada). The amount of lipids per sample was normalized to the protein concentration measured by the bicinchoninic acid assay.   51 3.2.5! Statistical methodologies Linear models were used to perform differential expression analysis using the Linear Models for Microarray and RNA-Seq data89 (limma) R-library. Statistical comparisons included comparing ERs and DRs at pre- and post-challenge. The post-challenge comparison used expression levels that had been normalized to the baseline expression levels (post/pre) in order to reflect molecular changes during the allergen inhalation challenge. The Benjamini-Hochberg False Discovery Rate (BH-FDR) was used to correct for multiple hypothesis testing91. Linear regression was used to test the association between immune cell frequencies and expression levels of cell-specific genes. Cell counts and all combinations of cell ratios (T, B, Treg and Th17) were also compared using linear regression. All statistical analyses were performed using the statistical computing program R95 (version 3.0.1).  3.2.5.1! Multivariate methods 3.2.5.1.1! Regularized Canonical Correlation Analysis (RCCA) RCCA maximizes the correlation between two high-dimensional datasets128 (p>>n) where variables contributing to this correlation from both datasets can be identified. Regularization of the correlation matrix for each dataset was performed by adding a multiple ($) of the identity matrix, where $1 and $2 were determined using a grid search using the mixOmics R library129.  3.2.5.1.2! Partial Least Squares (PLS) Similar to RCCA, PLS maximizes the correlation between two high-dimensional datasets X and Y of sizes nxp and nxq (where n is the number of subjects and p and q are the number of variables in each dataset respectively). However, PLS does not require regularization; instead it  52 reduces both datasets to components (linear combination of original variables) and maximizes the correlation between them. The PLS method96 in the mixOmics R-library was used to correlate two datasets corresponding to cell-specific genes.  3.2.5.2! Pathway analysis Ingenuity Pathway Analysis (IPA) was used to determine the top biological functions and canonical pathways in the list of differentially expressed genes. The filtering criteria were set to lung for organ and immune cells for cells. IPA uses Fisher’s exact test to identify enriched pathways and BH-FDR to adjust for multiple hypothesis testing.  3.3! Results 3.3.1! Differentially expressed genes and metabolites between isolated early and dual responders Table 3.1 displays the demographics for the 14 study participants (eight ERs and six DRs), that participated in the cat allergen inhalation challenge. Both groups had similar age distributions and met the initial percent drop in forced expiratory volume in one second (FEV1) criteria; a drop of greater than 20% (Figure 3.1). During the late phase (between 3 to 7 hours), the maximum percent drop in FEV1 in the DRs (Mean ± SD, -21.3±7.9%) was approximately 4 times greater than that in ERs (-5.1±4.2%).  53   Figure 3.1 Lung function measurements during the allergen inhalation challenge. Drop in FEV1 at regular intervals over time in ERs and DRs. Both groups exhibited an early response whereas the late phase response was significantly different between ERs and DRs. The LOESS (locally weighted scatterplot smoothing) curve was fitted to each group separately, surrounded by a 95% confidence interval.   At pre-challenge, absolute neutrophil counts were significantly greater in ERs compared to DRs (p=0.04). Figure 3.2 indicates that at pre-challenge, the relative levels of neutrophils were significantly greater in ERs compared to DRs (p=0.01), whereas relative levels of lymphocytes were significantly reduced in ERs compared to DRs (p=0.02). No significant differences in cell counts were identified post-challenge (normalized to pre-challenge levels).  Early Responder Dual Responder●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● ●●●●●●●●●●●● ●●●●●●●●●●●●●●●●●●●●●●● ●●●●●●●●●●●Blood draw●●●●●●●●●●●●●●● ●●●●●●●●●●●●●●●●● ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●Blood draw−60−40−2000 2 4 6 0 2 4 6Time (hours)Percent drop in FEV 1 54   Figure 3.2 Complete blood counts. Platelet counts and relative immune cell frequencies. At pre-challenge the percentage of neutrophils was significantly elevated in ERs whereas the percentage of lymphocytes was significantly elevated in DRs (p<0.05). Box and whiskers plots were made using the 25th (Q1) and 75th (Q3) percentiles of the data, where the thicker black line corresponds to the median value. The upper whisker extends from Q3 to Q3+1.5xIQR, where IQR is the interquartile range (Q3-Q1). The lower whisker extends from Q1-1.5xIQR. All points beyond these whiskers were plotted as single points.  Human Gene ST 1.0 microarrays which interrogates 32,321 probe sets were used to profile whole blood gene expression. Pre-filtering for informative probe sets (see Chapter 3.2.4.1) resulted in 770 probe sets of which 497 annotated probe sets were retained for further analysis.  55 From the same subjects, the plasma was analyzed for 292 annotated and 201 unannotated metabolites; only the annotated metabolites were retained for downstream analyses.  At pre-challenge, 72 (11 over and 61 under-expressed) differentially expressed probe sets in DRs relative to ERs were identified at a BH-FDR of 10%, after adjusting for age and sex. Ingenuity Pathway Analysis (IPA) indicated connective tissue disorders, immunological disease, inflammatory disease and skeletal and muscular disorder as the top biological functions at a BH-FDR of 5%. Significant genes, such as interleukin 8 receptor, alpha (CXCR1), nicotinamide phosphoribosyltransferase (NAMPT), c-c chemokine receptor type 7 (CCR7) and lymphoid enhancer-binding factor 1 (LEF1) have previously been shown to be abundantly expressed by neutrophils and lymphocytes. In fact, the differences in gene expression between ERs and DRs at pre-challenge were consistent (directionality) with the differences in cell-type frequencies (Figure 3.2). Eight metabolites were found differentially expressed between ERs and DRs at pre-challenge (p<0.05, BH-FDR>10%). Plasma levels of six metabolites including bilirubin, 4-vinyl phenol sulphate, 2-arachidonoylglycerophosphocholine, methionine, N-acetylglycine and malate were significantly elevated in DRs compared to ERs. Methyl palmitate and mannose were significantly under-expressed in DRs relative to ERs.  Twenty-five probe sets were differentially expressed post-challenge (normalized to pre-challenge levels) at a BH-FDR of 10%, after adjusting for age and sex (Appendix B.1). Six of these probe sets were up-regulated, and 4 probe sets were down-regulated in both ERs and DRs after challenge. The gene encoding RPL9 (ribosomal protein L9) was up-regulated in ERs and down-regulated in DRs at post-challenge. The remaining 14 genes were down-regulated in ERs and up-regulated in DRs after challenge. IPA revealed lipid metabolism (BH-FDR<5%) and linoleic acid metabolism (p<0.05) as the top biological function and canonical pathway,  56 respectively. Figure 3.3 shows a network of the differentially expressed genes consisting of the interleukin 5 receptor alpha (IL5RA), fatty acid desaturase 2 (FADS2), arachidonate 15-lipoxygenase (ALOX15), and cold shock domain protein A (CSDA).   Figure 3.3 Gene network. Network analysis of differentially expressed genes at a BH-FDR <10% between ERs and DRs in response to allergen challenge. Dash lines indicate indirect relationships, whereas solid lines indicate direct relationships. These interactions are based on known/predicted curated interactions in the IPA database. The red colors indicate differentially expressed genes whereas the gray colour corresponds to genes that were not differentially expressed. ALOX15 was down-regulated in both ERs and DRs (stronger down-regulation in ERs),  57 IL5RA was down-regulated in both ERs and DRs (stronger down-regulation in ERs), and FADS2 and CSDA were down-regulated in ERs and up-regulated in DRs.  Eleven metabolites were differentially expressed (p<0.05, BH-FDR>10%) post-challenge (levels normalized to pre-challenge levels) (Appendix B.1). Four of these metabolites were up-regulated whereas 2 metabolites were down-regulated after challenge in both ERs and DRs. The metabolite 1-pentadecanoylglycerophosphocholine was up-regulated in ERs and down-regulated in DRs after challenge. There were 4 metabolites that decreased in ERs and increased in DRs after challenge. The list of 11 differentially expressed metabolites was mainly enriched with lipid and amino acid metabolism pathways. Lipids were either lysolipids such as 1-pentadecanoylglycerophosphocholine, 2-arachidonoylglycerophosphocholine, 1-linoleoylglycerophosphocholine or sterols/steroids such as andro steroid monosulfate 1, cortisol and 7-alpha-hydroxy-3-oxo-4-cholestenoate (7-Hoca). The amino acid metabolism pathway consisted of metabolites such as 4-hydroxyphenylacetate, 3-methyl-2-oxobutyrate and cysteine. The polypeptide bradykinin, hydroxy-pro (3) increased in DRs and decreased in ERs in response to allergen challenge.  3.3.2! Gene-metabolite networks in ERs and DRs in response to allergen challenge Regularized Canonical Correlation Analysis (RCCA) was used to highlight clusters of correlated genes and metabolites in ERs and DRs separately. The ratio between pre and post gene expression levels (post/pre) was calculated for the 25 and 11 differentially expressed genes and metabolites for 14 responders resulting in 2 datasets; #G25,14 (25 significant genes for 14 responders) and #M11,14 (11significant metabolites for 14 responders). Figure 3.4 displays the correlations between #G and #M independently for ERs (Figure 3.4A; #G25,8 and #M11,8) and  58 DRs (Figure 3.4B; #G25,6 and #M11,6) using a correlation cut-off of 0.5. RCCA indicated greater numbers of correlated (r > 0.5) clusters in DRs compared to ERs. In dual responders (B; dotted circle), 2-arachidonoylglycerophosphocholine linked key enzymes (FADS2 and ALOX15) involved in lipid metabolism and IL5RA from the gene network in Figure 3.3.   59    Figure 3.4 Network plots highlighting the correlation between !G and !M for isolated early and dual responders. A. Gene-metabolite clusters for 8 isolated early responders (!G25,8 and !M11,8) and B. Gene-metabolite clusters for 6 dual responders (!G25,6 and !M11,6) using 25 differentially expressed genes (BH-FDR<10%) and 11 significant metabolites (p<0.05, BH-FDR>10%) between !ERs and !DRs. A correlation coefficient cut-off of 0.5 was applied to both networks. These bipartite networks connect genes (yellow circles) and metabolites (white rectangles) using both positive (solid red lines) and negative (dashed blue lines) correlations. All gene-metabolite correlations were calculated using the ratio (post/pre ) in gene and metabolite expression in ERs and DRs separately. A. Isolated Early Responders B. Dual Responders  60 3.3.3! !-6 and !-3 polyunsaturated fatty acids Both free and phospholipid forms of arachidonic acid (AA) and docosahexaenoic acid (DHA) were measured in a validation cohort of 19 (8 ERs and 11 DRs) study participants challenged with various allergens including cat, grass and ragweed (Table 3.2). The free fatty acid form is the fatty acid bound to albumin or on the surface of lipoproteins and usually refers to the metabolized lipid either from the hydrolysis of phospholipids or other complex lipids. The phospholipid form (membrane bound) is another structure which combines two fatty acids, usually at the sn-2 position in AA or DHA. Lipid levels were compared between ERs and DRs at post-challenge (normalized to pre-challenge) using a linear model adjusting for age and sex. At post-challenge, neither forms of AA were significant whereas only the free form of DHA was significant between ERs and DRs (p = 0.033), decreasing in ERs post-challenge compared to baseline levels whereas little change occurred in DRs (Figure 3.5).   61  Figure 3.5 Levels of free docosahexaenoic acid in the plasma of early and dual responders undergoing allergen inhalation challenge. At post-challenge (normalized to pre-challenge levels), DHA was differentially expressed (p = 0.033) between ERs and DRs. Levels of DHA decreased in isolated early responders (Early) from pre to post-challenge whereas no consistent change was observed in dual responders (Dual) following allergen challenge.   3.3.4! Th17/Treg ratio is associated with the late phase asthmatic response Sum of T and B cell frequencies obtained using the methylation assays strongly correlated (Spearman r = 0.95) with the lymphocyte frequencies obtained using a hematolyzer (Appendix 6.4B.2). T cell, B cell and Th17 cell counts were significantly positively correlated with the genes targeted in epigenetic cell counting in both the microarray (Figure 3.6; top row) and ●●P=0.0330.900.951.001.05Dual EarlyResponsePost−challenge levels (normalized to pre−challenge levels)Response●●DualEarly 62 NanoString (Figure 3.6; bottom row) platforms. Treg cell counts were correlated with FOXP3 gene expression measured using NanoString but not microarrays, suggesting higher resolution of the NanoString platform (Figure 3.6, red points).   Figure 3.6 Scatter plots of immune cells quantified using DNA methylation analysis with the corresponding cell-specific gene expression profiles. x-axis: relative cell-type frequencies of T, B, Treg and Th17 cells in whole blood; y-axis: a) top row: gene (CD3D, CD3G, CD79A, CD79A, FOXP3, and IL17A) expression intensities measured using microarrays and b) bottom row: gene expression counts measured using nCounter Elements from NanoString for the same set of samples. All cell-type specific genes showed strong correlation with their corresponding cell-types with the exception of FOXP3 expression vs. Treg counts (red points).   Allergen inhalation did not significantly change T cell, B cell, Treg cell and Th17 cell counts in either ERs or DRs. In addition, comparing the change in cell counts in ERs with the change in cell counts in DRs (ΔER vs. ΔDR), no significant cell-types were identified. Next, the ratios between all cell-type pairs were analyzed and only the Th17/Treg ratio was significant (p=0.03). Figure 3.7 shows that the Th17/Treg ratio increased in DRs compared to ERs, from pre to post- 63 challenge. The Th17/Treg ratio did not change from pre to post-challenge in ERs (net change = 0.006 ± 0.09), whereas the Th17/Treg ratio increased in DRs (net change = 0.28 ± 0.03).   Figure 3.7 Change in the Th17/Treg ratio in early and dual responders from pre to post-challenge. Th17/Treg ratio in ERs (left panel) and DRs (middle panel) at pre- and post-challenge. The change in the Th17/Treg ratio (post-pre) in ERs and DRs (right panel). Boxplots depict data from the 25th to 75th percentiles with the median value is represented by solid black points. The dashed lines connect the minimum and maximum points to the boxplots, whereas the solid lines connect paired samples (pre and post) of the same subject.  3.3.5! Association of gene-expression profiles with T helper 17 (Th17) cells, T regulatory (Treg) cells and the Th17/Treg ratio A multiple linear regression model (limma) was used to identify genes whose expression levels correlated with the frequencies of specific cell-types independent of changes in the frequencies of other cell-types. Ten genes were positively correlated with Th17 cells and 99 genes were positively correlated with Tregs at a BH-FDR of 10%, with no overlapping genes between the two  64 gene lists. Th17 genes included KIR2DS2, TAGLN, C14orf37, KRTAP13-3, SAP30, KIR2DS4, LAIR2, FLJ30679, RORC and KIR2DL2. The 99 Treg genes were enriched (BH-FDR = 5%) for 27 pathways, including many relevant regulatory pathways such as IL-2 regulation of translation, Regulation of telomere length and cellular immortalization, Regulation of T cell function by CTLA-4. Partial Least Squares (PLS) was used to determine the correlation between the set of 10 Th17 genes and the set of 99 Treg genes. Figure 3.8 depicts the results of PLS analysis using a correlation circle (see Chapter 2.3.3.2 on how to interpret a correlation circle or Gonzalez et al.130 for complete details on graphical outputs of PLS). Figure 3.8 shows that the Th17 genes were inversely correlated with Treg genes (angle greater than 90°).   65  Figure 3.8 Correlation circle depicting the strength of correlation between Treg genes (red) and Th17 genes (blue) with their respective latent variables (Comp 1 and Comp 2). Genes within each ellipse (cell-type) are positively correlated with each other, but inversely correlated with genes in the other cell-type.   13 genes significantly correlated with Th17/Treg ratio using limma (BH-FDR = 5%). Interestingly, 7 genes (KIR3DL1, LAIR2, KIR2DS2, KIR2DL2, CD226, KIR2DS4, KIR2DS1) belong to the leukocyte receptor complex located on chromosome 19q13.4, and were shown to  66 be positively correlated except CD226. Of the four genes profiled using NanoString, only CD226 and KIR2DS4 successfully replicated (Figure 3.9). The top-listed transcriptional network in GeneGo network analysis for the 13 significant genes included regulatory functions in immune responses.   Figure 3.9 Scatter plots of genes significantly correlated with the Th17/Treg ratio. x-axis: Th17/Treg ratio; y-axis: gene (CD226, KIR2DS4, KIR3DL1 and LAIR2) expression measured using a) top row: microarrays and b) NanoString nCounter Elements platform using the same set of individuals. The associations of both CD226 and KIR2DS4 expression with the Th17/Treg ratio successfully replicated (p<0.001) whereas the association of LAIR2 expression with the Th17/Treg ratio did not (p=0.43). The association between KIR3DL1 expression and the Th17/Treg ratio approached statistical significance (p=0.09).  3.4! Discussion Asthma is a complex disease consisting of multiple sub-phenotypes such that current therapies perform well in some phenotypes compared to others4. Therefore, well characterized study cohorts are required to identify new molecular targets of asthma. Despite the limited sample size  67 of the present study, many potentially confounding factors have been accounted for through careful selection of study participants. The participants were selected from a homogenous group of non-smoking individuals with stable, mild, atopic asthma and free of other lung diseases. Detailed characterization of the participants was based on the development of allergen-induced bronchoconstriction, and airway hyperresponsiveness to methacholine. Although airway samples are commonly used to study the pathobiology of respiratory disease, such samples are obtained using invasive procedures.  In the present study, we demonstrate the utility of applying ‘omics’ based approaches to peripheral blood in order to discriminate allergen-induced early from dual asthmatic responders (ERs from DRs). Although many genes were differentially expressed between ERs and DRs prior to allergen inhalation challenge, this may be influenced by significant differences in cellular frequencies between ERs and DRs. On the other hand, the cellular frequencies of ERs and DRs moved in parallel from pre to post-challenge, thus comparing post-challenge levels (normalized to pre-challenge levels) was deemed less likely to be influenced by differences in the major immune cell-type proportions. Therefore, greater focus was directed towards identifying significant molecular differences between ERs and DRs in response to the allergen inhalation challenge. These molecular targets may serve as effective biomarkers for the diagnostic monitoring of asthmatic responses in clinical trials or as therapeutics targets for the attenuation of the late phase asthmatic response.  Differentially expressed genes at post-challenge were enriched for lipid metabolism and linoleic acid metabolism pathways and included genes such as FADS2, ALOX15 as well as IL5RA, suggesting that these expression differences may be originating from eosinophils. EMR1, an eosinophil-specific receptor131 was also differentially expressed between ERs and DRs at  68 post-challenge. The mechanism of action of IL-5 on eosinophil survival involves the arachidonate pathway where phosphorylation of cytosolic phospholipase A2 " induces up-regulation of arachidonic acid levels in eosinophils132. Allergen inhalation challenge stimulates these pathways in circulating blood eosinophils132. FADS2 is a key enzyme in the biogenesis of arachidonic acid and docosahexaenoic acid133. ALOX15 can produce pro-inflammatory mediators such as eoxins134 and anti-inflammatory mediators such as lipoxins from arachidonic acid135, as well as pro-resolving mediators such as resolvins and protectins from docosahexaenoic acid134,136. Previous studies analyzing lipid mediators in bronchoalveolar lavage fluid have identified significant differences in ALOX15 metabolites between mild asthmatics and controls137,138. The latter study indicated higher levels of ALOX15 metabolites in mild allergic asthmatics compared to healthy controls, and this difference increased following allergen challenge. Thus, blood is a useful medium to depict lung inflammation since divergent oxylipin profiles can be observed in individuals with mild disease.  Lipid metabolism was also enriched in the list of significant metabolites at post-challenge. Cortisol, an immunosuppressive and anti-inflammatory molecule, decreased in ERs and increased in DRs after challenge. Cortisol bound to carrier proteins in the blood can be released at the site of inflammation by the catalytic activity of neutrophil elastase139. This result may suggest that the uptake of plasma cortisol in ERs leads to reduced inflammation whereas this mechanism is impaired in DRs. Many lysolipids which are important components of cellular membranes were differentially regulated between ERs and DRs after challenge. The release and subsequent metabolism of these lipids into pro- or anti-inflammatory mediators may dictate the associated clinical response. Bradykinin, hydroxy-pro(3) was differentially expressed between ERs and DRs at post-challenge; increased in DRs and decreased in ERs from pre- to post- 69 challenge. This result is consistent with the well-established role of bradykinin in asthma as a bronchoconstrictor and vasodilator140. In Chapter 2.3.3.2, Kininogen (KN1), which is a precursor of bradykinin was shown to decrease in DRs and increase in ERs after challenge (see Figure 2.4C). Therefore, plasma metabolomics reinforces the transcriptomics analysis and also identifies other biologically relevant molecules.  Although each data set (transcriptomics and metabolomics) is useful in identifying differences between ERs and DRs, considering each data set independently may result in some loss of information as both data sets were obtained from the same individuals. Regularized Canonical Correlation Analysis identified interesting patterns between differentially expressed genes and metabolites in ERs and DRs at post-challenge. Genes encoding for enzymes that can metabolize (ALOX15) or produce (FADS2) arachidonic acid were negatively correlated with the arachidonic acid lysolipid (2-arachidonoylglycerophosphocholine) in DRs after challenge. Although this network is biologically plausible, experimental work to validate the interactions amongst these molecules in peripheral blood is beyond the scope of this project. Interestingly, in the validation cohort of 19 study participants (8 ERs and 11 DRs), docosahexaenoic acid instead of arachidonic acid was differentially expressed between ERs and DRs at post-challenge. The free form of DHA was down-regulated in ERs post-challenge compared to pre-challenge whereas little change occurred in DRs from pre- to post-challenge. It may be possible that DHA is being taken up and metabolized in ERs leading to the resolution of inflammation136,138,141 whereas no such effect occurs in DRs within the two hour time frame. This pro- and anti-inflammatory axis was further evaluated using immune cell frequencies of T helper 17 (Th17) and T regulatory (Treg) cells. Although Th17 and Treg cells arise from a common precursor cell142 they have opposing inflammatory roles which has been  70 demonstrated in the context of autoimmune disease143, infection144, and allergic airway inflammation145. A potential Th17/Treg homeostatic imbalance in peripheral blood may exist between isolated early and dual asthmatic responders (ERs and DRs) undergoing allergen inhalation challenge. Lymphocyte subsets (T, B, Th17 and Treg cells) measured using DNA methylation analysis, were strongly correlated with their corresponding cell-specific gene expression profiles as measured by microarrays. Technical replication using nCounter Elements from NanoString, a higher resolution platform indicated that FOXP3 expression was indeed correlated with Treg cell counts. As a marker for human Tregs, however, FOXP3 expression is of doubtful value, due to its transient expression in activated non-regulatory effector T cells124. In addition, other cell-surface markers such as CD127 or CD45RA have been used to isolate FOXP3+ Treg cell populations with high efficiency146,147. Epigenetic enumeration of Treg cells in the present study has been shown to positively correlate with CD4+CD25+CD127lo, and CD4+CD25+CD127lo FOXP3+ and thus are truly representative of suppressive Tregs148.  The percentage of Treg cells did not significantly change in either ERs or DRs, two hours post-challenge. Previous studies have also not shown significant changes in Treg cells in peripheral blood in DRs undergoing allergen inhalation challenge72,149. This may be due to many factors such as the time of the post-challenge blood draw, the cell-surface markers used to isolate the Treg cells as well as the small sample sizes (n = 6-11) used in these studies. Similarly, the percentage of Th17 cells also did not significantly change in ERs or DRs after allergen challenge. Th17 cells have been shown to be increased 7 and 24 hours post-challenge in both ERs and DRs and the increase in DRs was greater than in ERs 24 hours post-challenge150. Th17 cell counts as well as the concentrations of IL-17 and IL-22, have also been shown to be increased with the severity of allergic asthma151. Genes significantly positively correlated with  71 Th17 cells included RORC, the transcription factor involved in Th17 differentiation, whereas genes significantly positively correlated with Treg genes were enriched for regulatory functions. Furthermore, Th17 and Treg cell associated genes were inversely correlated with each other, further implicating the phenotypic roles of these cell-types in allergic asthma.  Although neither cell-type significantly (p>0.05) changed pre to post-challenge, the change in the Th17/Treg ratio from pre to post-challenge significantly differed between ERs and DRs (p = 0.03). The Th17/Treg ratio increased in DRs whereas little change occurred in ERs after challenge. The increase in the Th17/Treg ratio in DRs was driven by an increase in the number of Th17 cells and a decrease in the number of Treg cells due to allergen inhalation challenge. A possible mechanism of Th17/Treg imbalance was suggested by the genes that were correlated with Th17/Treg ratio. LRC on chromosome 19q13.4 encodes immunoglobulin super family receptors including killer immunoglobulin like receptors (KIRs) expressed on natural killer (NK) cells and some subsets of T cells152. Almost all LRC genes except CD226 were significantly positively correlated with the Th17/Treg ratio. This suggests that an increase in NK cells may be associated with the late phase response. Previous studies have shown increased peripheral blood NK cells in asthmatics compared to controls and inhibition of pulmonary eosinophils and CD3+ T cell infiltration after depletion of NK and NKT cells in mouse models153,154. These studies further implicate NK cells in allergic asthma and offer support to the role of NK cells in the development of the late phase asthmatic response. However, this hypothesis needs to be validated in further work comprised of enumeration of NK cells in ERs and DRs using DNA methylation analysis. These studies implicate an imbalance between pro- and anti- inflammatory processes in allergic asthma, which may cause some individuals to become more susceptible to developing  72 the late phase asthmatic response or some individuals to be more protected from developing late phase responses.  3.5! Limitations An inherent limitation of transcriptional profiling studies using blood is that changes in gene expression may be due to differences in cell-type frequencies between groups. Frequencies of white blood cell counts may also influence plasma lysolipid levels and their metabolism can influence subsequent clinical phenotypes. Although the use of multiple datasets uncovered common biological processes, validation of candidate targets such as arachidonic acid was unsuccessful. This may be due to differences in the types of assay used to detect metabolites in the discovery and validation cohorts. Measuring the specific arachidonic acid lysolipid (2-arachidonoylglycerophosphocholine) may have been a more appropriate validation. Other limiting factors of the validation cohort apart from the small sample size include the use of various allergens and incomplete data on blood cell counts and differentials. The potency of different allergens is variable155 and can dictate the type of response an individual may experience; for instance, exposure to house dust mite results in a greater drop in FEV1 during the late asthmatic response compared to grass pollen156. Another limitation of the present study is that only a limited number of cell-types were studied using DNA methylation analysis, whereas quantification of a wide array of cell-types such as Th1, Th2, and Th9 cells would provide deeper biological insights into the mechanisms of allergic asthmatic responses. Unfortunately, DNA methylation assays for these additional cell-types had not been developed at the time. Fluorescence activated cell sorting may also be used but introduces additional sources of variability, which were avoided using DNA methylation analysis. Lastly, all integrative methods used in this section cannot incorporate more than two  73 datasets, thus an integration of cell counts, gene and metabolite profiles was not possible. This issue is addressed in Chapter 5 which focuses on the development and application of a multi-omic biomarker algorithm for the integration of more than two omics datasets.   3.6! Conclusion Although this study does not directly evaluate the impact of allergen inhalation on the airways, the downstream effects (in blood) of this lung insult can provide significant information towards further understanding the molecular mechanisms underlying the early and late asthmatic responses. This study has shown that careful selection of study participants is effective in detecting molecular differences between ERs and DRs. The use of multiple data sets from the same individuals can identify common biological functions and mechanisms and integration of these data sets may narrow down the number of generated hypotheses. Similar studies with larger sample sizes may reveal novel mechanisms and targets for the pathogenesis and treatment of the late phase asthmatic response.    74 Chapter 4:!Blood biomarker panels of the late phase asthmatic response 4.1! Introduction The late phase asthmatic response shares many characteristics with chronic asthma such as prolonged airway contraction, persistent airway inflammation, mucus hypersecretion and airway remodeling28,157. The late response also involves cellular infiltration of immune cells, however in isolated early responders (ERs) this influx is smaller compared to dual responders (DRs)158. Although inhaled corticosteroids are effective in abolishing the late phase response159, they do not alter the natural course of the disease160. Furthermore, airway inflammation (e.g. presence of epithelial shedding, T cells and Eosinophil Cationic Protein in mucosal biopsies) is present in subjects with mild disease and worsens in subjects with persistent asthma161. Given the heterogeneity and variability of the disease within an individual and over time, biomarkers that can help risk stratify individuals who display early indications of chronic, persistent asthma may help improve diagnosis, response to therapy and prevent unnecessary treatment. Biomarkers may also help uncover novel biological mechanisms that may serve as new treatment opportunities.  According to the Biomarkers, EndpointS, and other Tools (BEST) Resource developed by the Food and Drug Administration-National Institutes of Health (FDA-NIH), a susceptibility/risk biomarker is a “biomarker that indicates the potential for developing a disease or sensitivity to an exposure in an individual without clinically apparent disease”162. Although induced sputum is the gold standard for phenotyping asthma, it is an invasive and extensive procedure such that research is now starting to focus on noninvasive methods such as urine and blood. For example, Fractional Nitric Oxide (FENO) is a biomarker of airway inflammation and can be detected in exhaled breath condensate163. However, given the increasing number of conflicting studies and large variation of FENO values in the general population59, the use of  75 FENO alone may not be sufficient as a risk biomarker for asthma164. Periostin is a serum biomarker that correlates with eosinophilia in the airways60 and has been shown to improve response to anti-IL-13 antibody therapy. IL-13, produced by T helper type 2 (Th2) cells simulates the airway epithelium to release periostin42 which has roles in tissue remodeling in response to injury165. However, periostin is increased in many inflammatory diseases such as cancer, and myocardial and lung injuries and thus systemic levels of periostin may be attributed to comorbidities of asthma such as atopic dermatitis and rhinosinusistis165,166. Many studies have shown the association of serum periostin and blood eosinophil levels with sputum eosinophilia suggesting the use of peripheral blood to assess airway inflammation60,167. Others have replicated these results for blood eosinophils but not periostin168 but this may be due to the type of periostin assay used, type of hematolyzer used and differences in asthma cohorts with respect to asthma severity166. Nevertheless, these studies demonstrate the power of less-invasive methods for the detection of biomarkers of asthma. Biomarker panels have been shown to improve the predictive performance over single biomarkers and have been applied in various diseases (e.g. respiratory diseases, cardiac transplantation) and using diverse biological mediums (e.g. plasma, bronchial biopsies)169–173. This section of the thesis pertains to the identification of biomarker panels (combinations of single mediators) using gene expression in blood that can identify asthmatic individuals who will develop the late phase response prior to and after allergen inhalation challenge. Following the Institute of Medicine (IOM) guidelines for the development of clinical biomarker tests174, biomarker panels were identified in a discovery cohort with rigorous clinical phenotypic characterization. The biomarker panels were developed using RNA-sequencing data where the computational model performances were assessed using cross-validation. Then the identified  76 biomarker candidates were transferred to the more clinically relevant NanoString platform and the biomarker panel formulas were locked down. Lastly the performance of these biomarker panels were assessed using an independent external validation cohort.  4.2! Materials and methods 4.2.1! Cohorts 4.2.1.1! Discovery cohort A total of 36 subjects underwent allergen inhalation challenges and were classified into 15 ERs and 21 DRs (Table 4.1).   Table 4.1 Subject demographics of the discovery cohort Clinical variable Isolated Early responders (ERs) n=15 Dual responders (DRs) n=21 p-value Percent Female 67% 67%  Weight (kg) 70.2±14.8 73.0±16.3 0.47 Height (cm) 168.3±8.4 168.4±9.3 0.98 Age (years) 27.1±8.0 32.0±13.1 0.07 Baseline FEV1 (L/s) 3.4±0.8 3.2±0.8 0.44 % drop in FEV1 during the early phase (EAR) -35.2±9.9 -35.4±10.0 0.93 % drop in FEV1 during the late phase (LAR) -7.5±4.0 -28.7±15.1 1.2x10-10 Allergen induced shifta 1.7±1.7 3.4±1.9 0.001 Allergen (# of individuals) Birch 0 0  Cat 8 5 Fungus 0 2 Grass 3 2 HDM 4 8 Horse 0 1 Ragweed 0 3  77 Clinical variable Isolated Early responders (ERs) n=15 Dual responders (DRs) n=21 p-value Site (# of individuals) Laval 10 8  McMaster 5 12 UBC 0 1 a[PC20]pre/[PC20]post Note: summary statistics are expressed as Mean±SD. Grass: Grass mix, Orchard Grass, Timothy grass HDM: House dust mite, Dermatophagoides (D) Farinae, D. Pteronysinnus  4.2.1.2! Validation cohort For external validation, 45 subjects underwent allergen inhalation challenges and were classified into 9 ERs and 36 DRs (Table 4.2). Given the limited number of subjects remaining for validation purposes, 11/14 of the subjects from the discovery cohort in Chapter 3 (Table 3.1) were assigned to the validation cohort. Importantly, there was no overlap in the subjects between the discovery and validation cohorts.  Table 4.2 Subject demographics of the validation cohort Clinical variable Isolated Early responders (ERs) n=9 Dual responders (DRs) n=36 p-value Percent Female 56% 42%  Weight (kg) 70.2±11.2 73.3±14.5 0.59 Height (cm) 167.9±10.3 173.0±8.0 0.13 Age (years) 33.4±8.2 29.1±10.1 0.24 Baseline FEV1 (L/s) 3.0±0.5 3.5±0.7 0.048 % drop in FEV1 during the early phase (EAR) -30.6±10.3 -34.9±9.0 0.22 % drop in FEV1 during the late phase (LAR) -6.4±4.5 -22.5±8.7 3.1x10-16 Allergen induced shifta 1.2±0.8 3.0±1.8 0.02  78 Clinical variable Isolated Early responders (ERs) n=9 Dual responders (DRs) n=36 p-value Allergen (# of individuals) Birch 1 0  Cat 6 14 Fungus 0 0 Grass 2 7 HDM 0 12 Horse 0 1 Ragweed 0 2 Site (# of individuals) Laval 6 10  McMaster 1 19 UBC 2 7 a[PC20]pre/[PC20]post Note: summary statistics are expressed as Mean±SD. Grass: Grass mix, Orchard Grass, Timothy grass HDM: House dust mite, Dermatophagoides (D) Farinae, D. Pteronysinnus  Data were available for individuals who underwent repeat challenges and some did not elicit the same response upon repeat allergen challenge (possibly due to factors such as the allergen dose, allergen specificity, etc.). These individuals, termed “flippers”, were all labeled DRs due to their susceptibility to the late phase response (Appendix 6.4C.1).  In order to assess the expression of biomarker panel transcripts in healthy controls, four healthy individuals [two males (ages 22 and 29) and two females (ages 25 and 34)] were recruited at the Centre for Heart Lung Innovation, St. Paul’s Hospital, Vancouver. Blood samples were collected at 9 am and 12 pm for all four individuals.   79 4.2.2! Experimental techniques 4.2.2.1! Blood collection and processing Blood samples were collected immediately prior to (pre) and 2-3 hours after (post) allergen inhalation challenge using standard operating protocols at each participating Centre. At each time point blood was collected into EDTA tubes (3mL of blood) and PAXgene Blood RNA tubes (PreAnalytiX – Qiagen/BD, Valencia, CA, USA). Complete blood cell counts and differentials were obtained using a hemotalyzer [Cell Dyn 3700 System (Abbott Diagnostics, IL, USA)]. The EDTA blood tubes were then used to prepare the buffy coat, plasma and concentrated erythrocyte fractions by centrifugation at 500 x g for 10 minutes at room temperature. 2.5mL of blood was collected in PAXgene Blood RNA tubes which contain an additive that lyses red blood cells and stabilizes intracellular RNA preventing both degradation and changes in gene expression profiling due to sample handling. These samples were kept at -80°C prior to shipment to the Tebbutt laboratory in Vancouver, Canada.  4.2.2.2! RNA extraction After overnight thawing of PAXgene tubes, total RNA was purified using the RNA extraction kits such as PAXgene Blood RNA kit (PreAnalytiX-Qiagen, Germany) and PAXgene Blood miRNA Kit (PreAnalytiX-Qiagen, Germany) using 5mL of the PAXgene solution (~2.5mL of blood + 6.4 mL of stabilizing reagent). RNA concentrations were determined using NanoDrop 8000 Spectrophotometer (Thermo Scientific, Wilmington, DE, USA). RNA quality was assessed using the Agilent 2100 Bioanalyzer following the RNA 6000 Nano kit protocol (Agilent Technologies, Santa Clara, CA, USA).   80 4.2.2.3! Transcriptomics 4.2.2.3.1! RNA Sequencing (RNA-Seq) Total RNA was purified and sent to Génome Québec (Centre d’Innovation Genome Quebec et Université McGill) for whole transcriptome sequencing (8 samples per lane). Quality control (Nano Chip Bioanalyzer and Nanodrop) for each sample was repeated by Genome Quebec prior to sequencing. External RNA Control Consortium (ERCC) spike-in controls (92 sequences) were added to all samples. rRNA/globin-depleted stranded cDNA libraries were sequenced using an Illumina HiSeqTM 2000 sequencing system as 100 bp paired end reads.  4.2.2.3.2! nCounter Elements TagSets 100 base pairs (bp) of oligonucleotide probes were designed for all biomarker candidates using the nDesign portal and with the help of the bioinformatics team at NanoString Technologies. For unknown targets, the sequences were selected based on coverage plots of RNA-Seq data and provided to NanoString Technologies. NanoString Technologies provided the nCounter Elements TagSets which contains fluorescently labeled reporter tags (sequence of six spots of color specific for each specific mRNA transcript target) and a biotinylated universal capture tag175. During hybridization the reporter and capture probes bind to target-specific probes, probe A (50bp), and probe B (50bp), both of which bind to the single stranded RNA transcript target sequence. After several wash steps, this Tag Complex is immobilized to the glass cartridge and scanned allowing digital quantification of RNA sequences82,175.  Probe A and Probe B pairs for all biomarker candidates as well as both Probe A pools and Probe B pools were ordered from Integrated DNA Technologies, Inc. (IDT, Iowa, USA) (pools consists of all probes A or B for all biomarker candidates). Briefly, 30X Working Probe  81 Pools were created for a set number of assays resulting in a 8.3x fold dilution of the master Probe A and B stocks. The final concentration each probe in the 30X Working Probe A and B Pool was 0.6nM and 3nM, respectively. For a single run of 12 samples, 130µL of the hybridization buffer was added to 65µl of the TagSet followed by 65µL of the Extension TagSet. Then 13µl of the 30X Working Probe A Pool was added. After briefly inverting and spinning down of the TagSet tube 13µl of the 30X Working Probe B Pool was added. Lastly 39µl of RNAase-free water was added bringing the total volume of the master mix for 13 assays (dead volume of one extra assay) to 325µl (25µl per assay). Hybridization reactions were performed in sets of 12 using a strip of 12 tubes. 25µl of the master mix was added to each tube in addition to 5µl (100 ng) of the RNA sample. After 16 hours hybridization at 67°C using the Bio-Rad PCR system, the strip of tubes was placed in the nCounter Prep Station for purification and immobilization of Tag Complexes to the flow cell. After 3 hours of high sensitivity preparation, the flow cell was scanned using the nCounter Digital Analyzer under the maximum field of view (max FOV) setting (555 FOVs).  4.2.3! Data analysis 4.2.3.1! RNA-Seq data analysis All RNA-Sequencing files passed quality control metrics based on FastQC standards. Initially the first 12 bases of all 100 bp reads were trimmed and all left reads (R1) files were concatenated into one file and all right end reads (R2) were concatenated into another file using all samples. Of note is that for the purposes of the de novo assembly additional RNA-Seq data from asthmatic subjects from repeat challenges was included. The left and right concatenated files were used by the Trinity software176 (version r20131110) to construct a de novo assembly of the blood  82 transcriptome (Reverse forward RF-stranded library type was specified). The abundance estimates of Trinity contigs were estimated using RSEM177 (RNA-Seq by Expectation Maximization, version 1.2.11) using the Bowtie aligner.  A similar approach was used for genome-guided assemblies. However, after advice from Illumina on how to increase alignment rates, Seqtk (version 1.0) was used to trim the first 5 bases and the last 25 bases of all 100 bp paired end reads (this was not repeated for the Trinity de novo assembly). The University of California, Santa Cruz (UCSC) gene transfer format (GTF) file [which contained the genomic positions for Coding sequences (CDS)] and the list of known UCSC gene-isoforms3, was used to extract transcript sequences from the 2013 human reference genome4 (GRCh38 build) in order to build a reference transcriptome for RSEM. The abundance estimates of UCSC gene and gene-isoforms were estimated using RSEM (version 1.2.19) using the Bowtie2 (version 2.2.4) aligner. RSEM estimates the abundance of gene-isoforms and then sums up the isoforms of a given gene to determine the abundance of genes. 89,357 UCSC gene-isoforms corresponded to 42,465 genes whereas 258,403 Trinity “isoforms” corresponded to 212,373 Trinity “genes”. The Trinity dataset at the “isoform” sequence level was used for downstream analyses such that, probes-specific sequences could be designed for the validation phase.  STAR (Spliced Transcripts Alignment to a Reference, version 2.5.0a), a fast RNA-Seq alignment software178 was also used to align the paired end reads to the human genome using annotations from the GENCODE project179 (GENCODE release v215), which include protein coding genes, transcribed variants, long non-coding RNAs, and pseudogenes. Feature counts180                                                 3 http://genome.ucsc.edu/cgi-bin/hgTables?hgsid=419083647_CPaaYnAZklf85tfa5qclmzQszSKh  4 http://hgdownload.cse.ucsc.edu/goldenPath/hg38/chromosomes  5 http://www.gencodegenes.org/releases/reference_releases.html   83 in the Subread/Rsubread package (version 1.5.0) was used to estimate the abundance of 60,155 Ensembl transcripts as well the 92 ERCC control sequences.  UCSC and Ensembl identifiers were annotated using the biomaRt R library (version 2.26.0). Trinity contigs were annotated using Trinotate: Transcriptome Functional Annotation and Analysis6, which uses BLASTX to query nucleotide sequences (translated in protein sequences) against protein databases such as SwissProt and UniPort. For selected Trinity contigs that were transferred to the NanoString platform, the 100bp probe sequences were also queried against the human reference genome (GRCh38.p2) using BLASTN7. The count data for all mRNA datasets were normalized to log2-counts per million (log2 cpm)181 as follows:  where Xcounts is a p (variables) x n (samples) matrix of transcript counts and lib.size is sum of all counts in each sample. Alignment and mapping statistics for all four RNA-Seq datasets, UCSC genes, UCSC gene-isoforms, Ensembl and Trinity are provided in Appendix 6.4C.2. The 92 ERCC spike-ins were used to establish a lower limit of detection (Appendix 6.4C.3).  Three house-keeping transcripts were identified from each of the following datasets: UCSC genes, Ensembl and Trinity datasets. The transcripts were first ranked based on their coefficient of variation and the top 50 genes were selected from each dataset independently. The top three most stable (little variation in expression across all 36 pre and 36 post samples) of the 50 transcripts per dataset were identified using the geNorm algorithm182 in the NormqPCR R-library (version 1.16.0).                                                 6 https://trinotate.github.io/  7 http://blast.ncbi.nlm.nih.gov/Blast.cgi  Xnorm = log2Xcounts + 0.5( )Tlib.size+1( )*106⎛⎝⎜⎞⎠⎟ 84  4.2.3.2! NanoString data quality control and normalization All NanoString data were assessed for various quality control (QC) metrics183 such as the number of fields of view, binding density, linearity of positive control spike-ins and the lower limit of detection (see Appendix 6.4C.4 for a complete list of QC parameters).  NanoString uses six positive control ERCC spike-in controls corresponding to six different concentrations in the 30µL hybridization; 128fM, 32fM, 8fM, 2fM, 0.5fM and 0.125fM in order to assess the technical performance of the assays. Data normalization was performed independently in the discovery, validation and control cohorts. The geometric mean of all six positive controls for each sample was calculated. Then, the mean of all sample geometric means was divided by each sample’s geometric mean in order to determine a normalization factor for each sample. The expression data for each sample was multiplied by the corresponding sample’s normalization factor, thus normalizing for assay-to-assay variability. Technical variability was assessed by running replicates (aliquots) of the same RNA sample across cartridges and across time (same day and different days). Since two samples were collected from each of the four control subjects, the average of the replicates was used in the statistical analyses.  4.2.4! Statistical methodologies 4.2.4.1! Classification algorithms Elastic net184 (Enet) and random forest185 (Rf), two commonly used classification methods172,186–188 were used to identify biomarker panels of the late phase asthmatic response for each RNA-Seq dataset. Enet is a penalized regularization method that performs both shrinkage of regression coefficients (increases predictive performance) and variable selection (selects a limited number  85 of variables from high-dimensional datasets) resulting in a parsimonious model. This is achieved by finding # (regression coefficients) that minimizes:  where the elastic net penalty is . The elastic net penalty contains the " (between 0 and 1) parameter which controls the size (# of variables with non-zero coefficients) of the biomarker panel; where "=0 corresponds to a model where all variables have non-zero coefficients and " = 1 corresponds to a model with many zero coefficients and a few non-zero coefficients (selected variables). The tuning parameter $ controls the strength of the penalty and is estimated using cross-validation (see Chapter 4.2.4.2). Random forest is an ensemble method which combines the predictions of many tree classifiers. Tree classifiers are predictive models where the leaves (nodes) are phenotypic groups and the branches are made by splitting variables at particular values that best split the observations into their respective phenotypic groups. For the present analysis, 500 tree classifiers were built from 500 bootstrap (with replacement) datasets. Each tree was built by splitting randomly selected m variables: , p = number of variables in each split with no pruning. However, variables are ranked based on the Gini importance score which reflects how often a given variable is selected for a split and its ability to split samples into their respective classes. For unseen samples, the predictions of all 500 trees are averaged to achieve a consensus prediction. In order to keep the number of selected variables by Enet and Rf the same, for a given Enet panel size k, k top ranked variables based on the importance score y − Xβ 2 + λP(α )P(α ) = 1−α2 β 22 +α β 1m = p 86 were selected for Rf. Enet and Rf were implemented using the glmnet (version 2.0-1) and randomForest (version 4.6-12) R libraries.  4.2.4.2! Deep cross-validation Deep cross-validation (k x M-fold) was used to assess the classification performance of the biomarker panels. Briefly, the data was split into M folds, 1/M served as the test set and (M-1)/M served as the training set. For each fold, differential expression analysis was performed using the training set and the top ranked 500 transcripts (with the smallest p-values) were used to build the classification model. Each classification model was then used to predict the probabilities of the subjects in the held-out test set. This process was repeated until all folds had been used as a test set (M times). The result was a list of probabilities (likelihood of being a dual responder) for all subjects that were used to calculate the area under the receiver operating curve (AUC), a measure of classification performance. The AUC measures how well the biomarker panel separates two phenotypic groups of subjects. An AUC of 1 would depict a perfect blood test, whereas an AUC of 0.5 depicts a blood test that randomly allocates subjects into the two phenotypic groups. The M-fold cross-validation is repeated k times. When M = N, also called a leave-one-out cross-validation, each subject is treated as a test set.  4.2.4.3! Classification performance measures For both the discovery and validation phase, an AUC cut-off of 70% was used as per recommended guidelines for clinical biomarker implementation189. The Youden’s index190 was used to determine the optimal cut-off point where both the sensitivity and specificity are maximized, such that subjects with predicted probabilities below this threshold were classified as  87 ERs and above this threshold were classified as DRs. This cut-off was then used to calculate other classification indices such as the sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV). Sensitivity was defined as the proportion of true DRs the biomarker panels classified correctly whereas specificity was defined as the proportion of true ERs the biomarker panels classified correctly. PPV was defined as the fraction of predicted DRs that were true DRs, whereas NPV was defined as the fraction of predicted ERs that were true ERs. The PPV and NPV depend on the prevalence of the disease to which the biomarker test will be applied to; increasing disease prevalence increases the PPV but decreases the NPV. Since the biomarker test will be applied to the allergic asthmatic population in general where the prevalence of DRs is 60%22, this prevalence was used in the calculation of the PPV and NPV (, where prev=prevalence, sens=sensitivity and spec=specificity).  4.2.5! Gene-set enrichment analysis Tissue enrichment analysis was performed using the sear R library8 (version 0.1), which uses a hypergeometric test to test for over-representation (enrichment) of various cell-specific genes in the input gene list. Pathway enrichment was performed using Enrichr191, a web based application that allows the user to upload a list of gene symbols. The gene list is then compared to curated lists of canonical pathways/molecular processes from various databases [Kyoto Encyclopedia of Genes and Genomes (KEGG), Biocarta, Reactome, and the Encyclopedia of DNA Elements (ENCODE)                                                 8 https://github.com/cashoes/sear  PPV = prev* sensprev* sens + (1− prev)* spec and NPV =(1− prev)* specprev*(1− sens)+ (1− prev)* spec 88 Transcription Factor Chromatin Immunoprecipitation (ChiP-Seq) database] and tests for enrichment using a Fisher’s exact test.  4.3! Results 4.3.1! Changes in cellular frequencies and gene expression between ERs and DRs  4.3.1.1! Cell-count changes Complete blood cell counts were collected for a subset of the individuals, n=32/36 individuals (13ERs and 19DRs) at pre-challenge and n=26/36 (11ERs and 15DRs) individuals at post-challenge). Figure 4.1 depicts the levels of total leukocytes in ERs and DRs prior to and 2 hours after allergen inhalation challenge. Although the leukocyte counts appear elevated in ERs compared to DRs prior to allergen challenge, the difference is not statistically significant (p=0.23). In addition, changes in leukocyte levels due to allergen inhalation challenge are not statistically different between ERs and DRs (p=0.50).   Figure 4.1 Absolute leukocyte counts. Total leukocyte counts increased in both responder groups 2 hours after allergen inhalation challenge. The Mean±SD are plotted for both responder groups at the pre- and post-challenge time points.  ●●●●56789Pre PostTimeLeukocytes x109  Cells/LResponse●●ERDR 89 In addition, analyzing the five major immune cell-types in blood (neutrophils, lymphocytes, monocytes, eosinophils and basophils) did not yield any significant changes at either pre- or post (post minus pre expression) challenge time points (Figure 4.2). Only monocytes approached statistical significance (p=0.09) comparing ERs and DRs at pre-challenge; elevated in DRs compared to ERs.  Figure 4.2 Relative frequencies of immune cells in the blood. No cell-type was significantly different between early and dual responders at both pre-challenge and post-challenge (normalized to pre-challenge). The Mean±SD are plotted for both responder groups at the pre- and post-challenge time points.  4.3.1.2! Changes in gene expression Below 14 attomoles/µL or 3 log2 cpm, the linearity between the ERCC spike-in and RNA-Seq count data was diminished (Appendix 6.4C.3), therefore transcripts below this threshold in any of the 72 samples (36 pre- and 36 post) were removed prior to downstream analyses. Pre-filtering of low abundant transcripts resulted in 6,598 UCSC genes, 6,078 UCSC gene-isoforms, 7,518 Ensembl transcripts and 5,227 Trinity contigs. Principal component analysis on the joint dataset (25,421 transcripts) did not uncover any effects due to confounding variables such as allergen challenge date, sex, site, and allergen (significant overlap between 95% confidence ellipses, see Appendix 6.4C.5). ●●●●●●●●●●●● ●●●●●●●Neutrophils Lymphocytes Monocytes Eosinophils Basophils0.500.550.600.650.700.200.250.300.350.400.060.070.080.090.100.000.040.080.0000.0050.010Pre Post Pre Post Pre Post Pre Post Pre PostTimePercentage of cellsResponse●●ERDR 90 Differential expression using limma identified greater numbers of differentially expressed genes between ERs and DRs at the pre-challenge compared to post-challenge (post-challenge expression was scaled to pre-challenge expression, post minus pre-challenge expression). Figure 4.3 shows that the discriminatory signal was consistent across datasets. At a BH-FDR threshold of 25%, 2,143 UCSC genes, 1,520 UCSC gene-isoforms, 1,192 Ensembl transcripts and 986 Trinity contigs were significantly different between ERs and DRs at pre-challenge.  Figure 4.3 Transcripts ranked based on BH-FDR for each dataset and time point. The pre-challenge time point identified a greater number of differentially expressed genes between ERs and DRs compared to post-challenge at a BH-FDR=25%.  UCSC genes UCSC gene−isoforms Ensembl Trinity255075100255075100Pre−challengePost−challenge10 1000 10 1000 10 1000 10 1000Number of transcriptsBH−FDR 91 Annotating the 5,841 specific IDs of all datasets to gene-symbols resulted in 3,025 unique gene-symbols. Tissue enrichment of these gene symbols indicated significant enrichment for various immune compartments of blood and low enrichment for other tissue-type such as the spleen, tonsils, and central nervous system (Figure 4.4).   Figure 4.4 Tissue enrichment of differential expressed genes at pre-challenge. Enrichment of various tissue-types in the list of significant genes between ERs and DRs at pre-challenge. Amongst the various cell-types, the gene list was highly expressed in various immune cells most specifically in different T cell compartments (BH-FDR < 0.00001% or –log10(BH-FDR) > 5).  05101520CNSThymusLymph nodeSpleenTonsilsB cells PreIB cells PreIIB cells ProB cells CD19+B cells immatureB cells cord blood CD34+Macrophage LPS 4hDC LPS 6hDC LPS 48hDC BAFF+Mast cell IgEDerived Macrophage 16hMacrophageMast cellMyeloid CD33+Monocyte CD14+DC ImmatureDCNK CD56+BloodRBCNeutrophilsThymic CD34+Thymic CD4+CD8+CD3−Thymic CD4+CD8−CD3−Thymic CD34+CD38+ CD1A−Thymic CD4+CD8+CD3+T cells BAFF+Thymic CD34+CD38+ CD1A+T cells cord blood CD34+Thymic SP CD4+ T cellsT cells gammadeltaT cells CD57+ Th1Th2TregThymic SP CD8+ T cellsPeripheral CD8+ T cellsPeripheral naive CD4+ T cellsT cells effector memoryT cells central memoryTonsils CD4+ T cells−log10 (BH−FDR)CNSTHYMUSTONSILSSPLEENLYMPHBCELLSMYELOIDNKCELLSBLOODRBCSNEUTROPHILSTCELLSTH1CELLSTH2CELLSTREGSCD8TCELLSCD4TCELLS 92 Figure 4.4 indicates that the differentially expressed genes can mainly be attributed to different T cell subsets.  Of the 5,841 differentially expressed transcripts between ERs and DRs at pre-challenge, 948 were up-regulated in DRs whereas 4,893 were down-regulated in DRs compared to ERs. Gene-set enrichment analysis using Enrichr identified two highly ranked (p < 0.01) KEGG pathways, leukocyte transendothelial migration and apoptosis that overlapped between the up-regulated and down-regulated gene lists (Figure 4.5). The leukocyte transendothelial migration pathway consisted of 31 differentially expressed genes, 24 genes down-regulated in DRs and 7 genes up-regulated in DRs. In the apoptosis pathway 19 genes were down-regulated in DRs and 7 genes were up-regulated in DRs at pre-challenge.    93  Figure 4.5 Top 20 pathways of the up and down-regulated genes in DRs compared to ERs at pre-challenge. Overlapping bars represent top ranked pathways in both up and down-regulated gene lists.  egf signaling pathwayregulation of eif2proteasome complexil−7 signal transductiondouble stranded rna induced gene expressioninternal ribosome entry pathwayrac1 cell motility signaling pathwaybasic mechanisms of sumoylationskeletal muscle hypertrophy is regulated via akt−mtor pathwaynfkb activation by nontypeable hemophilus influenzaehypoxia−inducible factor in the cardivascular systemrole of pi3k subunit p85 in regulation of actin organization and cell migrationfas signaling pathway (cd95)regulation of eif−4e and p70s6 kinaseceramide signaling pathwayeukaryotic protein translationp38 mapk signaling pathwayt cell receptor signaling pathwayctcf: first multivalent nuclear factorhiv−1 nef: negative effector of fas and tnfras signaling pathwaysignal dependent regulation of myogenesis by corepressor mitrmcalpain and friends in cell motilityrole of mef2d in t−cell apoptosisakt signaling pathwaymets affect on macrophage differentiationmapkinase signaling pathwayrole of nicotinic acetylcholine receptors in the regulation of apoptosisrole of erbb2 in signal transduction and oncologyhuman cytomegalovirus and map kinase pathwayschromatin remodeling by hswi/snf atp−dependent complexesrole of erk5 in neuronal survival pathwaycarm1 and regulation of the estrogen receptorrho cell motility signaling pathwaycontrol of skeletal myogenesis by hdac and calcium/calmodulin−dependent kinase (camk)erk and pi−3 kinase are necessary for collagen binding in corneal epitheliaucalpain and friends in cell spreadil−2 receptor beta chain in t cell activationhemoglobins chaperone0 3 6 9−log10 (P−value)Biocarta pathwaysprotein exportb cell receptor signaling pathwaytoll like receptor signaling pathwaymapk signaling pathwayepithelial cell signaling in helicobacter pylori infectionregulation of autophagypyruvate metabolismglycan structures biosynthesis 1basal transcription factorscell cyclecolorectal cancerchronic myeloid leukemiafolate biosynthesisproteasomen glycan biosynthesisubiquitin mediated proteolysispathogenic escherichia coli infection epecpathogenic escherichia coli infection ehecapoptosisleukocyte transendothelial migrationprostate cancererbb signaling pathwaynatural killer cell mediated cytotoxicityglyoxylate and dicarboxylate metabolismglycan structures degradationdorso ventral axis formationcitrate cyclefocal adhesionnotch signaling pathwaynon small cell lung cancerglycosaminoglycan degradationrenal cell carcinomavegf signaling pathwayregulation of actin cytoskeletonporphyrin and chlorophyll metabolismadherens junctionendometrial cancer0 2 4 6−log10 (P−value)KEGG pathwaysDirectionDown in DRsUp in DRs 94 Figure 4.5 indicates that T-cell related pathways in the Biocarta database were identified in both gene lists; up-regulated genes in DRs were enriched for IL-2 receptor beta chain in T cell activation whereas down-regulated genes in DRs were enriched for T cell receptor signaling. Interestingly, the top ranked pathway in the up-regulated genes in DRs was the hemoglobins chaperone pathway based on the Biocarta database.  4.3.2! Classification performances of pre-challenge biomarker panels Given that the discriminatory signal between ERs and DRs was stronger at pre-challenge (Figure 4.3), effort was allocated to identifying risk biomarker panels prior to allergen challenge. Screening biomarker panels that can identify asthmatic individuals who are more likely to develop dual asthmatic responses upon allergen inhalation challenge offer a number of advantages. Firstly, the allergen inhalation challenge is not required to identify dual responders, only one blood draw is required from each participant, and the screening blood test can be applied to the allergic asthmatic population in general.  The workflow from discovery to validation of predictive biomarker panels is shown in Figure 4.6. The four RNA-Seq datasets comparing 15 ERs and 21 DRs were analyzed in parallel in order to first, estimate the test performance of the pre-challenge classifiers, second, transfer a selected number of biomarker candidates to the NanoString platform for discovery of biomarker panels suited for this platform and lastly, validation of final biomarker panels in an external independent cohort.   95  Figure 4.6 Workflow of biomarker panel discovery and validation. Multiple datasets (genome guided and de novo assembly) were generated for the RNA-Seq data from 36 subjects of the discovery cohort. Biomarker candidates were identified from each RNA-Seq dataset and transferred to the NanoString Elements platform. Biomarker panels were re-identified for each dataset specifically and when all candidates were combined in 29 out of the 36 subjects. The test performance of each biomarker panel was assessed using an external independent cohort of 45 subjects. RNA$SequencingUCSCgenesUCSCgene$isoforms TrinityEnsembl200x5%fold* CVp=6,598 p=5,227p=7,518p=6,078Estimate*test*error* of*biomarker*panels*using*Cross%Validation* (CV)Frequently*occuring genes1000*panels200x5%fold* CV1000*panels200x5%fold* CV1000*panels200x5%fold* CV1000*panelsStability*Analysis:*select*final*biomarker*candidates* from*genes*that*occur* most*frequently* across*the*biomarker*panelsFrequently*occuring contigsFrequently*occuring transcriptsFrequently*occuring gene%isoformslarge*panelsmall*panel200x5%fold*CV200x5%fold*CV200x5%fold*CVDiscoveryRe$discoveryValidationsmall*panel small*panel small*panel small*panelDetermine?Optimal?Panel?Size?Lock?down?biomarker?panels?and?assess?performance?in?an?independent?validation?cohortUCSC?panel UCSC?gene?isoform?panelEnsembl?panel Trinity?panel All?panelnanoStringDesign?100bp? nanoString probe?sequences40?genes 36?gene$isoforms 49?transcripts 24??contigs⍺ =*0...⍺ =*1200x5%fold*CV200x5%fold*CV200x5%fold*CV⍺ =*0...⍺ =*1200x5%fold*CV200x5%fold*CV200x5%fold*CV⍺ =*0...⍺ =*1200x5%fold*CV200x5%fold*CV200x5%fold*CV⍺ =*0...⍺ =*1200x5%fold*CV200x5%fold*CV200x5%fold*CV⍺ =*0...⍺ =*1large*panel large*panel large*panel large*panelUCSC genes UCSC gene−isoforms Ensembl Trinity AllAUC = 0.56AUC = 0.51AUC = 0.67AUC = 0.67AUC = 0.53AUC = 0.54AUC = 0.71AUC = 0.67AUC = 0.62AUC = 0.6302550751000255075100EnetRf0 25 50 75 100 0 25 50 75 100 0 25 50 75 100 0 25 50 75 100 0 25 50 75 100100 − SpecificitySensitivity 96 In the discovery phase (Figure 4.6), the classification performance for the pre-challenge classifiers was based on the AUC for two commonly used classification methods, elastic net (Enet) and random forest (Rf) (see Section 4.2.4.1 for description of methods). The estimate of the test performance for all classifiers was on average greater than 60% (Figure 4.7). The Ensembl Enet classifiers on average had a performance of AUC = 69.4±7.6% whereas the Trinity Enet classifiers met the set threshold of an AUC of 70% at 70.4±6.7%.   Figure 4.7 Estimate of test performance as measured by the AUC using a 200x5-fold deep cross-validation. All datasets had an average AUC across the 200 iterations of 60% or greater, the UCSC gene classifiers with the weakest performance and the Trinity classifiers with the highest performance based on the AUC.  Since the test performances estimated for the pre-challenge classifiers either met or came close to the set threshold of AUC=70%, it was deemed suitable to proceed to validation in an independent cohort in order to confirm this discriminatory signal. ●●●●●●●●6070UCSC genesUCSC gene−isoformsEnsemblTrinityDatasetsAUCClassifier●●EnetRf 97  4.3.3! Biomarker candidate selection for screening classifiers Since Enet performs both shrinkage and selection (Section 4.2.4.1) the selection parameter (") was set to 0.9 for all RNA-Seq datasets resulting in panel sizes that ranged from 15.6±6.7 to 18.1±7.3 transcripts across the 1000 panels generated in the 200x5-fold deep cross-validation. For a given Enet panel size (k = # of transcripts in an Enet panel), the top ranked k transcripts based on the importance score were selected for the Rf panels since Rf does not perform variable selection. Transcripts across the 1000 panels were tallied for both the Enet and Rf panels for each dataset. The overlap between the top 200 ranked transcripts in the Enet and Rf was determined for each dataset and annotated, from which a subset of candidates was selected: 40 UCSC genes, 36 UCSC gene-isoforms, 49 Ensembl transcripts, and 24 Trinity contigs (Figure 4.8, also see Appendix 6.4C.6 for overlap between transcripts selected from each dataset).   98  Figure 4.8 MA plot of biomarker and house-keeping candidates. Relationship between fold-change and average expression for biomarker candidates and selected house-keeping genes. The selected house-keeping transcripts (within the dashed lines) have little change in expression between ERs and DRs.  4.3.4! House-keepers Nine house-keeping transcripts were identified across three datasets, UCSC genes, Ensembl and Trinity (Figure 4.8). However, since the Trinity house-keeping contigs were well above the expression of all other selected transcripts and may lead to saturation problems in the NanoString platform, they were removed resulting in the 6 house-keeping gene candidates; MED13, TOR1AIP2 and WAC (Ensembl dataset) and ARPC4, TMBIM6 and RHOA (UCSC gene dataset), all of which were transferred to the NanoString platform.  ●●●●●●●●●●●●●●●●●● ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● ● ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● ●●●●●●●●●●● ●●●●●●●●●●●●●●●●●●●●●●● ●●●●● ●●●House−keepers−1.0−0.50.00.54 6 8 10 12Average Expression log2 cpmlog2 fold−change (DR minus ER)Datasets●●●●UCSC genesUCSC gene−isoformsEnsemblTrinity 99 4.3.4.1! NanoString Elements probe design Six house-keeping probes as well as 87 probes (see Appendix 6.4C.7 for sequences of all transcripts) corresponding to the 149 biomarker candidates (due to gene redundancy across datasets, see Appendix 6.4C.6 for overlap between datasets) were designed. Transcripts with UCSC gene and Ensembl identifiers were annotated to their corresponding gene symbol. Based on the gene name, a probe was selected from the nDesign portal with the assumption that for each gene, the 100bp NanoString probe would have an expression similar to that of the much longer sequence of nucleotides from the RNA-Seq data.  Gene-isoforms were designed for 10/36 UCSC gene-isoforms; CD4_isoform, CNTNAP3_isoform, COPB1_isoform, IKBIP_isoform, MAP3K8_isoform, PELI1_isoform, PTAR1_isoform, TIA1_isoform, VPS13A_isoform, and ZNF609_isoform. For the remaining 26/36 gene-isoform candidates specific gene-isoform probes could not be designed due to high sequence similarity with other isoforms, therefore gene specific probes (targeting multiple isoforms) were selected (see Appendix 6.4C.7 for UCSC isoform-specific identifiers).  The Table in Appendix C.8 shows that 3/24 contig sequences were paralogs of a single contig which mapped to LYST. Two transcripts were designed for comp55647_c0_seq2, one mapping to a discriminatory region and the other to non-discriminatory region (Appendix C.10). Therefore, the trinity panel consisted for 23 unique contigs. 18/23 Trinity contigs had strong sequence identity with known genes (see Table in Appendix 6.4C.8) based on alignment to the GRCh38.p2 human genome reference assembly. For example, the Trinity contig comp56950_c0_seq1 (3,899 base pairs) was able to fully reconstruct the sequence of HCLS1 (Appendix 6.4C.9). Appendix 6.4C.9 also shows the mapping of the 100bp probe targeting HCLS1 that spans two exons. The remaining 5 of the 23 Trinity contigs mapped either to intronic  100 regions of known genes or to uncharacterized loci (Appendix 6.4C.10). Short sequences (>100bp) were selected based on visualizing the base coverage of each transcript and sent to NanoString technologies which selected 100bp from this sequence as the final probe sequence (Appendix 6.4C.11).   4.3.5! Rediscovery using NanoString Elements platform 4.3.5.1! Quality control and reproducibility assessment Probes for 87 transcripts were ordered from IDT and assayed in 29 out of 36 samples since RNA for some samples in the discovery cohort had been completely depleted. The test run of the custom Elements assays had extremely high binding densities for most samples (Figure 4.9). If the binding density is greater than 2 then many codes are ignored and this may affect quantification of the biomarker candidates. Figure 4.10 shows that this is mainly due to saturation by HBA2 transcripts, whose counts are reaching the maximum limit of 2 million counts per assay, whereas the sum of all other biomarker genes including control probes remains below 0.2 million counts.    101  Figure 4.9 Binding densities of samples in the test run of the Elements assay prior to and after attenuation of HBA2.   Given the extreme saturation bias of HBA2, an inactive probe192 was designed (same sequence as the HBA2 but no bound reporter probe) in order to achieve a 95% attenuation (see Appendix 1.6.4C.12 for sequence and calculations). Samples run with and without the inactive HBA2 probe indicated that the attenuation was in fact between 62-82%. Taken together with the high binding densities prior to HBA2 attenuation, it may be reasonable to conclude that the true HBA2 counts were well above the count limit of the NanoString assay. HBA2 not attenuated (test) Discovery cohort − HBA2 attenuated Validation Cohort − HBA2 attenuated●●●● ●●●●●●●●●●●●●●●●●●● ●●●●●●●●●● ●●●●●●●●●●●●● ●●● ●●●● ●●●●●●●●● ●●●●● ●●●●● ●●●●●●● ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●012E1_1E2_2E3_3E4_4E5_5E6_6E7_7E8_8E9_9E10_10E11_11E12_12E13_1E14_2E15_3E16_4E17_5E18_6E19_7E20_8E21_9E22_10E23_11E24_12E25_1E26_2E27_3E28_4E29_5E30_6E31_7E32_8E33_9E34_10E35_11E36_12E2_2E3_3E4_4E5_5E6_6E9_9E11_11E12_12E13_1E18_6E19_7E24_12E30_6E31_7E33_9E7_1E20_2E8_5E14_7E10_9E32_10E15_11E20_12E40_7E37_8E39_10E41_2E40_8E38_3E42_6E43_8E40_12E40_4E1_1E16_4E21_9E22_10E23_11E25_1E26_2E27_3E28_4E29_5E34_10E35_11E27_4E27_8V1_1V2_2V3_3V6_6V7_7V11_11V12_1V13_2V14_3V16_5V18_7V22_11V24_1V16_2V28_6V31_9V32_10V34_12V35_1V40_6V44_10V45_11V50_4V52_6V54_9V55_10V56_11V57_12V59_2V62_6V63_7V64_8V65_9V66_10SamplesBinding Density 102  Figure 4.10 Effects of HBA2 saturation on the custom elements assay. The total sum of all candidates measured per sample for the test, discovery and validation cohorts.  Spiking in the HBA2 inactive probe resulted in a significant reduction in the binding densities of the discovery and validation cohort samples (Figure 4.9). Figure 4.10 shows that not only did HBA2 counts significantly decrease after attenuation of HBA2, but also counts of other genes significantly increased.  Aliquots of a single RNA sample were run on different cartridges as part of the discovery cohort run (January 2016) and the validation cohort (February 2016). Figure 4.11 depicts excellent sample correlations (Spearman) ranging between 0.993-0.998 using log2 transformed raw data of all 87 biomarker candidates, suggesting minimal effects on expression profiles due to the experimental protocol. Test Discovery Cohort Validation Cohort●●●●100000200000500000100000015000002000000HBA2 other genes HBA2 other genes HBA2 other genesTotal counts for each assayattenuation of HBA2no attenuation of HBA2 103  Figure 4.11 Scatter plot matrix of 87 transcripts measured on aliquots of a single RNA sample across batches and across days. Excellent Spearman correlations between the technical replicates indicating minimal technical variability.  However, gene to gene (NanoString-RNA-Seq) correlations of the 87 transcripts ranged between -0.6 to 0.6. Figure 4.12 depicts a histogram of Spearman correlations of biomarker candidates measured on the RNA-Seq and NanoString platform. The Ensembl dataset (purple peaks) had 2016−01−166 8 12 160.998 0.9976 8 12 160.99346812160.995681216●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●2016−01−16 0.998 0.995 0.996●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●2016−01−16 0.9956812160.996681216●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●2016−02−10 0.9964 6 8 12 16●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●6 8 12 16●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●6 8 12 166812162016−02−11 104 many genes with very poor correlations across platforms, whereas the UCSC gene-isoforms and Trinity dataset had mostly positive correlations. The correlations for UCSC genes are scattered across the entire range of correlation values.  Figure 4.12 Histogram of gene-gene correlation of biomarker candidates between NanoString and RNA-Seq corresponding to the different datasets.  4.3.5.2! Optimal biomarker panel size  Since biomarker panels contain genes that work in conjunction with one another in order to predict the response-type, removal of genes with little concordance across platforms was not performed. Instead the optimal panel size (# of transcripts) for each dataset was identified using a 200x5-fold cross-validation repeated for different " values (0, 0.3, 0.5, 0.7, and 1), i.e. ranging from the largest panel to the smallest possible panel (see re-discovery phase in Figure 4.6), in order to identify the alpha value associated with the highest classification performance based on the AUC (Figure 4.13).  0.000.250.500.751.001.25−0.4 0.0 0.4Spearman correlation between RNASeq and nanoStringDensityDatasetUCSC genesUCSC gene−isoformsEnsemblTrinity 105  Figure 4.13 Classification performance of different panel sizes using the NanoString data subsetted according to the different RNA-Seq datasets. The 87 transcripts were split into 4 groups, each corresponding to the originating RNA-Seq dataset. “All transcripts” were all combined in order to identify a biomarker panel that combines different types of transcripts found on different RNA-Seq datasets. The panel size ranges from the largest panel size where all genes are in the panel ("=0) to the smallest possible panel ("=1).  For each panel in Figure 4.13, the classification performance drops as the panel size decreases, that is, the biomarker panel with all genes corresponding to each dataset results in the best classification performance. Therefore, the final biomarker panels that were locked down (fixed weights by fitting the model to all the data) prior to external validation in an independent cohort were the UCSC genes panel (40 transcripts), UCSC gene-isoforms panel (36 transcripts), Ensembl panel (49 transcripts), and the Trinity panel (23 transcripts) and lastly the All panel which contained all 87 transcripts. Of note is that given the high redundancy across the ●●●●●●●●●●4060801000 0.3 0.5 0.7 1alphaAUC (Mean+/−SD) 200x5−fold CVUCSC genes●● ●●●● ●● ●●4060801000 0.3 0.5 0.7 1alphaAUC (Mean+/−SD) 200x5−fold CVTrinity transcripts●●●●●●●●●●4060801000 0.3 0.5 0.7 1alphaAUC (Mean+/−SD) 200x5−fold CVUCSC geneIsoforms●●●●●●●●●●4060801000 0.3 0.5 0.7 1alphaAUC (Mean+/−SD) 200x5−fold CVall transcripts●●●●●●●●●●4060801000 0.3 0.5 0.7 1alphaAUC (Mean+/−SD) 200x5−fold CVEnsembl transcripts●●●●●●●6070UCSC genesUCSC gene−isoformsEnsemblTrinityDatasetsAUCClassifier●●EnetRf 106 biomarker panels, some transcripts were shared across panels (see Appendix 6.4C.6 for overlaps between the panels).  4.3.6! Validation in an independent cohort The five biomarker panels [UCSC genes, UCSC gene-isoforms, Ensembl, Trinity, and All (all transcripts combined)] were locked down (fixed weights of transcripts in the biomarker panels) by fitting the model to the entire discovery datasets containing 11 ERs and 18 DRs. An “off the shelf” test was performed using an external independent cohort of 9ERs and 36DRs for all biomarker panels and the AUC was determined (Figure 4.14). The AUCs of the Enet panels (left panel) were greater than the Rf panels (right panel) in general (Figure 4.14). The AUC of the UCSC genes and Ensembl biomarker panels were below 60% for both classification algorithms and hence these panels were deemed unsuccessful.   107  Figure 4.14 Classification performance measures in the validation cohort. Note the AUC corresponds to the validation cohort of 9ERs and 36DRs. The other classification performance measures such as the PPV, NPV, Sensitivity and Specificity were based on the cut-off determined using the Youden index.  The AUC for the UCSC gene-isoforms and Trinty panels was 68% and 72% respectively using the Enet classifier. Both panels were considered successfully validated given the AUC cut-off of 70%. Based on the Youden index (see Section 4.2.4.3 for details) the UCSC gene-isoforms Enet panel had a PPV and NPV of 80% and 55% whereas the Trinity Enet panel had a PPV and NPV of 89% and 59% respectively (see equations of classifier in Appendix C.12). Furthermore, the predictive performance of both biomarker panels was greater than that of random prediction when the class labels of the validation cohort were re-shuffled 100 times (Figure 4.15).   Enet Rf●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● ●●●●●●●●●●●●●●●●●●●●●406080100AUC Youden−cutoff PPV NPV Sensitivity Specificity AUC Youden−cutoff PPV NPV Sensitivity SpecificityPerformance MeasuresProbabilityBiomarker Panel●●●●●UCSC genesUCSC gene−isoformsEnsemblTrinityAll 108 UCSC gene-isoforms panel Trinity panel   Figure 4.15 Comparison of UCSC gene-isoforms and Trinity panel with random classifiers. After random shuffling of class labels of subjects in the validation cohort the UCSC gene-isoforms and Trinity panels had a random classification performance.  4.3.6.1! UCSC gene-isoforms biomarker panel The UCSC gene-isoforms biomarker panel consisted of 36 transcripts (Figure 4.16) where 13 out of 36 transcripts in the UCSC gene-isoforms panel were differentially expressed (BH-FDR<10%) in the combined discovery & validation cohort of 20 ERs and 54 DRs. Seven transcripts, PPP3R1, PABPC1, MBNL3, COPB1_isoform, PSMF1, CD4_isoform and NAPA were significantly over-expressed in DRs compared to ERs, whereas six transcripts CLEC4E, CNTNAP3_isoform, ABHD5, IL1R2, FPR2 and RGS2 was significantly under-expressed in DRs compared to ERs. Comparing the expression of these biomarker panel transcripts between healthy controls and ERs/DRs indicated that 28 of the 36 genes (all up-regulated) were differentially expressed between ERs and healthy controls and 26 of the 36 (all up-regulated) were differentially expressed between DRs and healthy controls (BH-FDR<10%). The UCSC genes-isoforms biomarker panel was significantly enriched with the Biocarta T cell receptor signaling pathway ●●●●●●●●● ●●●●●●●● ●●●●●●● ●● ●● ●●●●● ●●●●●●● ●● ●●● ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●AUC = 68%AUC = 51%02550751000 25 50 75 100100 − SpecificitySensitivityBiomarker Panel●●Random ClassifierUCSC gene−isoforms panel●●●●●●●●●●●● ●●●●●●●●●●● ● ●●●●●● ●● ●● ●●●●● ● ●●● ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●AUC = 72%AUC = 50%02550751000 25 50 75 100100 − SpecificitySensitivityBiomarker Panel●●Random ClassifierTrinity panel 109 (p=0.00043), consisting of five overlapping genes, NFKBIA, CD4, PPP3R1, CD8A and MAP3K8. Based on the ENCODE transcription factor ChiP-seq gene sets, the transcription factor CTCF regulated 13 biomarker gene candidates (CTSA, IKBIP, ATP8A1, IL1R2, ABHD5, NFKBIA, CD8A, PELI1, CD59, MAP3K8, TGFBI, MBNL3, and FAM8A1) in CD14 monocytes (BH-FDR<1%) in humans.    110  Figure 4.16 Fold-changes of transcripts of validated biomarker panels. Change in gene expression between ERs and DRs at pre-challenge for transcripts in the UCSC gene-isoforms and Trinity biomarker panels. *BH-FDR<10%, Transcript labels in red denote common genes between panels.MBNL3PSMF1PPP3R1CARM1COPB1_isoformCD4_isoformPABPC1HBA2NAPAPTPN18TGFBIF13A1VPS13A_isoformPELI1_isoformPTAR1_isoformSF3B1CTSAIKBIP_isoformCD59TIA1_isoformMBNL3PSMF1HBA2RAB5BCYTB_mitochondrionGenomeSF3B1TPP12_LOC101927568_intronCNTNAP3_isoformCLEC4EIL1R2RGS2FPR2ABHD5MMEFUT7NFKBIASEMA4DMAP3K8_isoformCD8AFAM8A1ZNF609_isoformSULT1A1ATP8A1TNFRSF10C_intronRGS2FPR2TNFRSF10C_intron.LOC254896_exonHCLS1FPR1_intronQKICASP8FNIP1SETX2a_LOC105378945_intronLYSTIFRD1_intronNFKB1CECR1* * ****** **** * ***** * *UCSC gene−isoforms panel Trinity panel−0.6−0.30.00.3log2 fold−change (+ up in DRs & − down in DRs) 111 4.3.6.2! Trinity biomarker panel The Trinity biomarker panel consisted of 23 transcripts (Figure 4.16) where 7 out of 23 transcripts were differentially expressed (BH-FDR<10%) in the combined discovery & validation cohort with four transcripts, MBNL3, PSMF1, RGS2 and FPR2 overlapping with the UCSC gene-isoforms panel. Comparing the expression of these biomarker panel transcripts between healthy controls and ERs/DRs indicated that 20 of the 23 genes (all up-regulated) were differentially expressed between ERs and healthy controls and 21 of the 23 (all up-regulated) were differentially expressed between DRs and healthy controls (BH-FDR<10%). Very few transcripts overlapped with curated gene-sets based on Enrichr analysis. Apoptosis (CASP8, TNFRSF10C, and NFKB1) and formyl peptide receptors bind formyl peptides and many other ligands (FPR1, and FPR2) were two of the top ranked pathways from the KEGG and Reactome databases (BH-FDR<1%). The transcription factor ETS1 was found to significantly (BH-FDR<1%) regulate 8 out of 23 biomarker panel candidates (SETX, RGS2, CASP8, IFRD1, HCLS1, TPP1, SF3B1, and NFKB1) in CH12-LX murine cell lines (ENCODE transcription factor ChiP-seq database).   4.3.7! Biomarker panels in response to allergen challenge Since the post-challenge samples of the validation cohort were not profiled using the NanoString platform, the post-challenge data from the discovery RNA-Seq dataset (15 ERs and 21 DRs) were used to determine whether the change in expression of biomarker panels retained their discriminatory potential after allergen challenge. The change in expression (post minus pre) was computed and the predictive potential of the UCSC gene-isoforms and Trinity panels was determined using a leave-one-out cross-validation scheme. The UCSC gene-isoforms panel had a  112 cross-validated AUC of 62% whereas the Trinity panel had a cross-validation AUC of 50% (Figure 4.17).   Figure 4.17 Predictive potential of the UCSC gene-isoforms and Trinity biomarker panels 2 hours after challenge. Using the RNA-Seq data, the post-challenge data was normalized to pre-challenge levels and the performance was estimated using a leave-one-out cross-validation.  4.4! Clinical implications The two biomarkers panels that validated (AUC~70%) had positive predictive values 20-29% higher than the prevalence (60%) of dual responders in the asthma population. Furthermore, the negative predictive values were 15-19% higher than the prevalence (40%) of early responders in the asthma population. These biomarker panels may be used as blood tests for the identification of at-risk asthmatic individuals that possess early indications of chronic asthma. Using baseline ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●0 20 40 60 80 100020406080100UCSC gene−isoforms100 − Specificity (%)Sensitivity (%)AUC =  62%●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●0 20 40 60 80 100020406080100Trinity100 − Specificity (%)Sensitivity (%)AUC =  50% 113 gene expression profiles from the general asthma population (given the inclusion/exclusion criteria), individuals can be screened for entry into clinical trials in order to examine the efficacy of new pharmacological interventions in attenuating the late response to allergen challenge. Furthermore, the UCSC gene-isoforms biomarker panel but not the Trinity panel retained a modest level of predictive power two hours after allergen inhalation challenge. Exploring the mechanistic link of these biomarker panel transcripts may provide greater insight into the mechanisms leading to the development of the late phase response and may ultimately lead to novel targets for asthma therapeutics.   4.5! Discussion Transcriptional profiling of blood identified significantly altered gene expression profiles in DRs compared to ERs prior to undergoing allergen inhalation challenge. These discriminatory signals were not sustained two hours after allergen inhalation challenge which may indicate similar immune responses during that time period. Therefore, effort was focused on the development and validation of susceptibility/risk biomarker panels that could significantly improve the identification of asthmatic individuals that would elicit the dual response compared to randomly screening the general asthma population using an allergen inhalation challenge. Of the four panels developed, two panels, the UCSC gene-isoforms and Trinity panels achieved an AUC of approximately 70% in an external validation cohort. The UCSC gene-isoforms panel, but not the Trinity panel, achieved a modest classification performance (AUC=62%) two hours post-challenge which may suggest that transcripts within this panel may play active roles in the development of the late phase asthmatic response. These results imply inherent differences in the  114 underlying molecular mechanisms that predisposes asthmatic individuals for the late phase asthmatic response.  Although the development of the risk biomarker panels was developed in asthmatic subjects only using baseline blood samples, the majority of transcripts in these panels were significantly under-expressed in healthy subjects. Combination of transcripts may represent biological processes that are dysregulated between groups of subjects193 and this may help increase their predictive power over single marker biomarkers58. Single mediators may be influenced by various disease related or independent factors whereas several mediators linked mechanistically might be more specific and robust to an underlying biological process. Although no mechanistic studies were performed in this thesis, biomarker panels with transcripts that belong to similar molecular pathways may support a mechanistic link. Taken together, the two biomarker panels (UCSC gene-isoforms and Trinity) implicated various biological pathways mainly in T cells [up- (down) regulation of CD4 (CD8A) in DRs compared to ERs].  Many transcripts in the biomarker panels indicated the activation of the nuclear factor (NF)-!B signaling pathway in dual responders suggestive of active pro-inflammatory responses. NF-!B is a ubiquitous transcription factor that induces the production of pro-inflammatory cytokines, chemokines, adhesion molecules and growth factors194. Coactivator-associated Arginine Methyltransferase 1 (CARM1), a cofactor of NF-!B195 and Pelino E3 Ubiquitin Protein Ligase 1 (PELI1) involved in the Toll Like Receptor and IL-1 signaling pathway which leads to the activation of NF-!B196 were up-regulated in DRs compared to ERs. The decoy receptor of IL-1, type 2 IL-1 receptor (IL-1R2) which binds IL-1197, was down-regulated in DRs, which indicates that during an immune response, more IL-1 can bind to IL-1R resulting in subsequent activation of pro-inflammatory pathways. The IL-1R antagonist (IL-1RA) Anakinra significantly  115 reduced airway neutrophilia in healthy subjects challenged with LPS as well as associated pro-inflammatory cytokines such as IL-1", IL-6 and IL-8198. Matrix Metallo-Endopeptidase (MME) which was down-regulated in DRs has been shown to be expressed on B cells and IL-4 producing follicular helper T cells199, where NF-!B activation leads to a down-regulation of MME mRNA and protein. NF-!B inhibitor (NFKB1) which was down-regulated in DRs binds to Mitogen-Activated Protein Kinase Kinase Kinase 8 (MAP3K8) which was up-regulated in DRs in order to prevent NF-!B activation200. NF-!B inhibitor alpha (NFKBIA), another inhibitor of NF-!B was down-regulated in DRs compared to ERs201. I!B kinase " (IKKB), up-regulated in DRs induces the activation of NF-!B by phosphorylating its inhibitor I!B leading to its subsequent degradation by the proteasome202. The caspase recruitment domain proteins B cell lymphoma 10 (Bcl10) and the paracaspase mucosa-associated lymphoid tissues 1 (Malt1) forms a complex domain that activates IKK in T cells and B cells after antigen receptor binding203. Mast cells from mice deficient in these proteins do not produce NF-!B induced proteins such TNF-# and IL-6 after stimulation of Fc$RI. Furthermore, the late response, but not the early response, was diminished in these animals suggesting that pro-inflammatory cytokines as a result of NF-!B activation lead to the development of the late phase response204. Caspase 8 (CASP8), a proapoptotic protease205 down-regulated in DRs, has been shown to regulate IL-1 cytokines inducing Th2 responses in mice leading to airway eosinophilia and Th2 differentiation but not of Th1 and Th17 cells206. Furthermore, tumor necrosis receptor superfamily member 10C (TNFRSF10C) which was down-regulated in DRs, is a decoy receptor that prevents apoptosis by competing with death receptors for tumor necrosis factor (TNF)-related apoptosis-inducing ligands and can be induced via a NF-!B dependent mechanism207. These results indicate an inherent heightened pro-inflammatory and anti-apoptotic phenotype in dual responders which  116 may explain their exaggerated response to allergen inhalation challenge. These results support an IgE-independent mechanism that occurs in the absence of mast cell degranulation.  Apart from a heighted inflammatory response, a dampened anti-inflammatory response was observed through reduced gene expression of the formyl peptide receptor family FPR1_intron and FPR2 in DRs compared to ERs. FPR1 and FPR2 are G-protein coupled receptors expressed by phagocytic cells such as monocytes, macrophages and dendritic cells that play roles in host defense against pathogen208. FPR2 binds lipoxin A4 (LXA4), an anti-inflammatory lipid metabolite that inhibits neutrophil chemotaxis and promotes phagocytic responses seen during the resolution of inflammation209. Introducing LXA4 in mice undergoing allergen challenge markedly reduced airway hyperresponsiveness, and airway inflammation210. FPR2 expression was decreased in induced sputum cells of children with severe asthma compared to intermittent asthma and healthy subjects211. Both receptors can also bind Annexin-1 (ANXA1)212,213, another anti-inflammatory molecule that reduces leukocyte adhesion and migration, phospholipase activity, prostaglandin an leukotriene production214. Glucocorticoids, the standard asthma therapy induces expression of ANXA1215 and is up-regulated in the bronchoalveolar lavage fluid of allergic asthmatic patients216. ANXA1-deficient mice challenged with ovalbumin exhibited higher levels of allergen-specific antibodies and increased airway hyperresponsiveness. Lower levels of receptor expression of FPR1 and FPR2 suggests a reduced capacity to bind anti-inflammatory ligands thus impairing the resolution of inflammation.  Apart from known genes, some uncharacterized transcripts that were constructed de novo were also present in the Trinity biomarker panel. These included intronic sequences in genes such as FPR1, IFRD1, TNFRSF10C and non-coding RNA (LOC254896, LOC101927568 and LOC105378945). A previous study indicated that ‘dark matter’ (unassigned functional role or  117 un-annotated) RNA represents a significant proportion of total RNA217 of sequenced reads in normal and neoplastic tissues. 50-65% of all non-ribosomal, non-mitochondrial RNA was ‘dark matter’ RNA suggesting a greater discovery space for identifying novel transcripts associated with disease. Non-coding RNA are known to regulate gene expression through mechanisms such as chromatin remodeling, post-transcriptional regulation and mRNA degradation218. For example, the large intergenic non-coding RNA (lincRNA) HOTAIR has been shown to be a strong predictor of metastases of breast cancer progression219. Prostate Cancer Antigen 3 (PCA3), a prostate tissue specific ncRNA that can be measured in the urine, is a stronger diagnostic biomarker for prostate cancer relative to the Prostate-Specific Antigen (PSA)220. Given the unexplored territory of ncRNAs in the context of asthma pathogenesis, the transcripts identified in the present study may just be the low hanging fruit given the depth of sequencing. The functional role of these transcripts at present remains unknown and future work will be required to delineate their biological functions. Given their reproducibility across platforms and cohorts these transcripts may not be technical artifacts of the data, but surrogates of the underlying biology.  The biomarker panels identified in this chapter may have a mechanistic link to underlying biology of the late phase asthmatic response, which may explain their predictive power in discriminating ERs and DRs in different cohorts. Using these panels in clinical practice may help screen mild asthmatic individuals who are susceptible to developing the late asthmatic response upon allergen exposure.   118 4.6! Limitations The variability from different sources such as subject heterogeneity, range of allergens used, and cellular heterogeneity is an inherent aspect of this study. The FEV1 profiles of the late phase asthmatic response (LAR) can be highly variable suggesting a more continuous phenotype of the late asthmatic response. A continuous variable such as the area under the FEV1 profile (between 3-7 hours) curve can also be used but would require significantly larger cohorts to detect significant associations. Using a binary cut-off with respect to the maximum drop in FEV1 during the LAR, subjects in the extreme tails of the distribution (% FEV1 drop in LAR) given the cohort, were selected to maximize chances of identifying a discriminatory signal. Certain allergens such as house dust mite and cat allergens are more prone (>75%)22 to inducing the late response, therefore ERs and DRs were matched as best as possible with respect to the allergen used for the inhalation challenge.  Cellular heterogeneity can also affect the results of this study such that the biomarker panels could represent differences in cell-type frequencies between responders rather than differences in cellular gene expression. However, no significant changes in cell-types frequencies of the major blood cells were identified. The results indicate changes in different regulatory processes (pro- and anti-inflammatory processes) as well as specificity to different cell-types (e.g. T cells) therefore the changes identified between ERs and DRs are most likely a combination of both differences in cellular frequencies and expression. The molecular signal between ERs and DRs, two hours post-challenge did not sustain its discriminatory power from the pre-challenge time point. Blood samples are collected two to three hours after a drop in FEV1 of 20% (early asthmatic response) has occurred after allergen inhalation. Therefore, this variability of the post-challenge blood draw may contribute to the dilution of the discriminatory  119 signal. Although we have shown differential gene expression between ERs and DRs, at two hours post-challenge in Chapter 3, a longitudinal study that can capture changes between responders over time would provide better insight into the ideal time point when the differences between responders is maximal or minimal.  Lastly, it may be argued that the independent cohort used in this study is not truly independent as all subjects in this cohort were not recruited after those from the discovery cohort. However, the present study is a multi-center study (three sites) with a greater proportion of samples from UBC used in the validation cohort compared to the discovery cohort.   4.7! Conclusion Blood biomarker panels can identify asthmatic individuals at risk of developing the late asthmatic response upon allergen inhalation. These panels consist of transcripts that display a stronger pro-inflammatory phenotype and a weaker anti-inflammatory phenotype in DRs. Several non-coding RNAs with uncharacterized functions were also present in these panels, but whether they represent novel regulatory functions remains to be determined. These panels may be used to assess allergic asthmatic subjects for early indications of chronic asthma and may ultimately lead to earlier diagnosis and improved asthma control.   120 Chapter 5:!Novel methods for the integration of multiple omic datasets for biomarker discovery  5.1! Introduction Data integration as defined by Ritchie et al.221 is the combination of multiple omic datasets using multivariate models that are predictive of complex traits or phenotypes. Data integration has typically been grouped into two categories221, concatenation-based112,113 and model-based integration222. Concatenation-based integration combines multiple datasets into one large dataset, which is used to predict a phenotype of interest. This approach favors the most discriminatory dataset, diluting the signal from the other datasets and also significantly increasing the number of variables. Model-based integration approaches such as ensemble classification construct a predictive model on each dataset separately and subsequently combine the model predictions. These approaches cannot account for relationships between datasets since predictive models are constructed independently for each dataset. Therefore, novel integrative methods for biomarker discovery are needed that can 1) explain the relationship between biological layers to improve insight into disease mechanisms, 2) be flexible in handling different data types and study designs such as repeated measures and 3) be easily interpretable and accessible to researchers. Such integrative methods will become the mainstay as technologies capable of simultaneously measuring multiple omics data-types such DNA-Encoded Antibody Libraries223 or the 3D biology NanoString platform224 become the norm.  The final section of this thesis pertains to the development of DIABLO (Data Integration Analysis for Biomarker discovery using a Latent variable approach for Omics studies), a multivariate classification method that identifies a discriminative subset of correlated variables that are predictive of multiple groups of subjects using multiple omic datasets. DIABLO was used  121 to conduct two studies: 1) obtain a holistic view of the molecular interactions in blood after allergen inhalation challenge and 2) develop a multi-omic biomarker panel that is predictive of an asthmatic individual’s response to allergen challenge. This chapter builds on the work of the previous two chapters using systems approaches in a discriminant analysis framework.  5.2! Materials and methods Two sets of datasets were used in this Chapter for two separate studies (Figure 5.1). Cell count, microarray, and metabolomics datasets generated in Chapter 3 were used to identify discriminative patterns between pre and post challenge across all asthmatic individuals (Study 1). Second, metabolite profiling was performed on the pre and post blood samples of 32 out of 36 subjects of the discovery cohort for whom RNA-Seq data were also available. Surrogate variables of cell-type frequencies were estimated using gene expression data (see Section 5.2.2). A multi-omic biomarker panel that was able to predict a subject’s response to challenge prior to allergen challenge was derived using these datasets (Study 2). The performance of this panel was also evaluated in post-challenge blood samples.   122  Figure 5.1 Workflow of the DIABLO analysis. In Study 1, DIABLO was used to identify multi-omic molecular changes between pre and post allergen challenge using datasets from Chapter 3. Metabolite profiling was performed on pre and post samples of 32/36 subjects from the discovery cohort in Chapter 4. Cell-type frequencies were estimated from a subset of the RNA-Seq data. A multi-omic biomarker panel was identified that was predictive of the late phase asthmatic response prior to allergen inhalation challenge. The performance of this panel was also assessed using post allergen challenge blood samples.  5.2.1! Gene expression profiling The filtered Ensembl dataset from Chapter 4.3.1.2 consisting of 7,518 transcripts, was used to determine a multi-omic biomarker panel using DIABLO.  5.2.2! Surrogate cell-type frequencies Cell marker genes were used to infer surrogate variables representing the frequencies of specific cell-types using the cellCODE R-library (version 0.99.0). cellCODE (Cell-type Computational Differential Expression) uses singular value decomposition (also termed eigengene summarization225) in order to estimate surrogate variables from sets of genes that are specific for DIABLOStudy&1Pre&vs.&Post Study&2ER&vs.&DRCellular&spaceTranscriptomics spaceMetabolomics&spaceComparison:&14&Pre&vs.&14&PostDatasets• 9&cellAtypes&(CBCs/Diffs)• Gene&expression&(microarrays)• Metabolite&profiling& (Metabolon)All&datasets&from&Chapter&3Comparison:&15&ERs&vs.&17&DRsDatasets• 84&cellAtypes&(estimated)*• Gene&expression&(RNAASeq)*• Metabolite&profiling& (Biocrates)*RNAASeq data&from&Chapter&4Group&1Group&2 123 certain cell-types226. Since cell-specific genes may be transcriptionally influenced by the phenotype itself, cellCODE removes this phenotypic group effect using a modified version of the two-step Surrogate Variable Analysis procedure227. Cell marker genes were obtained from 1) the IRIS (Immune Response In Silico)228 and 2) DMAP (Differentiation Map)229 datasets within the cellCODE R-library, 3) the LM22 signature matrix9 from Newman et al.230 (compiled data for 22 cell-types from 5 sources including the IRIS dataset) and 4) cell-specific genes annotated by NanoString Technologies as part of the nCounter Human PanCancer Immune Profiling Panel10. Many redundancies existed within the combined marker gene-sets from the above studies such as assigning different marker genes to the same cell-type albeit with a significant overlap in marker genes. Cell-types with less than 4 marker genes were excluded and at most 15 markers genes were considered per cell-type. The resulting cell marker matrix consisted of 789 genes corresponding to 84 cell-types (see Appendix D.1 for a correlation matrix of cell-types using the marker genes from the different studies). The Ensembl dataset (60,155 transcripts) from the RNA-Seq from Chapter 4.2.3.1 was used to obtain the surrogate cell-type frequencies for the 84 cell-types using cellCODE.  5.2.3! Metabolomics profiling kit The AbsoluteIDQTM p150 kit (Biocrates Life Sciences AG, Innsbruck, Austria) was used for absolute or relative quantification of 163 metabolites (41 acylcarnitines, 14 amino acids, 92 glycerophospholipids, 15 sphinolipids, and hexose) for both pre and post challenge samples of 32 out of 36 subjects (15ERs and 17DRs) of the discovery cohort in Chapter 4 (see Table 4.1). Briefly, 10 µl of the samples (plasma samples, blank controls of phosphate-buffered saline,                                                 9 https://cibersort.stanford.edu  10 http://www.nanostring.com/products/pancancer_immune   124 quality controls and an isotope-labeled internal standard) were pipetted onto the filter spots of a 96-well kit plate and dried under a nitrogen stream. For derivitization of amino acids, 20 µl of a 5% solution of phenylisothiocyanate was added. After drying again under a nitrogen stream, metabolites were extracted using 300 µl of a 5 mM ammonium acetate solution in methanol. The extracts were then analyzed using high performance chromatography coupled to a SCIEX 5500 QTrap®. Data analysis was performed using MetIDQTM software (Biocrates Life Sciences AG).  5.2.4! Pathway enrichment analysis InnateDB, a database that integrates publically available databases such as Reactome231, Integrating Network Objects with Hierarchies (INOH)232, BioCarta11, NetPath233, and KEGG12 as well as its own manually curated interactions234, was used for gene set enrichment analysis (called over-representation analysis by InnateDB). The associated web-interface13 was used for pathway over-representation using the hypergeometric test.  5.3! Results 5.3.1! Integrative classification method Classification algorithms a priori are not focused on incorporating biological information and therefore any derived discriminative markers (“biomarkers”) may not mechanistically link the underlying biology to the phenotype. To address this concern, DIABLO (Data Integration Analysis for Biomarker discovery using a Latent variable approach for Omics studies, see Appendix D.2 for details), an integrative classification method (Figure 5.2) was developed, which not only identifies subsets of discriminative molecules from each omic dataset, but also aims to more                                                 11 http://cgap.nci.nih.gov/Pathways/BioCarta_Pathways  12 http://www.genome.jp/kegg/pathway.html  13 http://allergen.innatedb.com   125 plausibly explain the correlation structure between them, assuming that correlation implies similar functional relationships235.  126  Figure 5.2 Integrative methods for biomarker discovery. Integrative classification methods for biomarker discovery. The concatenation method combines multiple omic datasets into one joint matrix which is then used to model a multi-group phenotype. The ensemble method builds a model for each omic dataset independently. For each new subject, predictions made by each model are averaged. DIABLO maximizes the sum of pairwise correlations between multiple connected datasets (user defined or computationally derived), where the phenotype vector is treated as a block dummy matrix. Similar to the ensemble method, the predictions made by the DIABLO model are averaged. X1 X5X3 X4X2 YX1X5X3X4X2YConcatenationEnsembleDIABLOMulti-omic classification5model Classifier5predictionConcatenation-based5classifierDIABLOEnsemble5classifiernew5subjectX1X2X4X3X5X1X2X4X3X5Average5predictionsAverage5predictionsX1X5X3X4X255omics datasetsnxp1nxp2nxp3nxp4nxp5PhenotypeYPhenotype51Phenotype52Phenotype53X1X5X3X4X2YSelect5designconnect5highly5correlated5datasets−1−0.8−0.6−0.4−0.200.20.40.60.81X1 X2 X3 X4 X5X1X2X3X4X510.280.880.560.2410.30.360.9210.870.1510.56 1X1X2X3X4X5 127 Briefly, DIABLO summarizes each omic dataset into latent variables which are linear combinations of the original variables. DIABLO then aims to maximize the correlation between latent variables of different omic datasets and the phenotype of interest. Datasets can be specified to be connected by the user either using 1) a priori knowledge, for instance, since miRNA regulate gene expression, these two datasets can be specified to be connected (correlated), or 2) the proposed method which determines the correlation between datasets such that dataset-pairs above a given correlation cut-off are assumed to be connected (Appendix D.3). DIABLO takes into account the connections between datasets which are summarized in a design matrix (Figure 5.2). DIABLO also performs variable selection on each omic dataset, retaining highly correlated and discriminatory variables in the multi-omic biomarker panel. Figure 5.2 compares DIABLO with common integrative methods such as the concatenation and ensemble-based approaches. Unlike the concatenation based approach, DIABLO allows equal contribution from each omics dataset. Instead of combining independent models as in the ensemble-based integration, DIABLO allows omics datasets to be correlated (through the design matrix) such that different omics datasets are able to influence each other’s predictions. In contrast to the concatenation and ensemble classification approaches which require the use of specific classification algorithms (Chapter 4.2.4.1), DIABLO itself is an integrative classification method that enables the integration of multiple omic datasets in order to predict a given phenotype.   128 5.3.2! Pipeline for biomarker discovery A biomarker pipeline (Figure 5.3) was developed in order to guide researchers on how to construct a multi-omic biomarker panel, assess its predictive performance, and generate visualizations in order to better interpret the outputs of the DIABLO classifier.  Figure 5.3 DIABLO biomarker pipeline. Multiple datasets from the same individuals with multiple (two or more) phenotypes are required to run a standard DIABLO pipeline as follows: 1) identify a multi-omic biomarker panel, 2) determine its classification performance using cross-validation and 3) use visuals to interpret phenotypic clusters, relationship between variables, heatmaps and pathway enrichment. InputPreprocessingMultivariate2analysisOutputsOmics datasets AimsmRNAmiRNACpGsProteins• Normalization• Filtering2of2featuresMulti?level2variance2decompositionaccount2for2repeated2measurementspathway2modulesPCA unsupervised2 analysismulti?omic biomarker2panelmolecular2datasetsClassifier2 performance Feature2selectionVisualizations• Sample2plots• Circos plots• Heatmaps• Pathway2enrichmentM?fold2cross2validationsample2clusteringrelationships2 between2omics datasetsBiological2assessmentMetabolomicsSet2design2matrixModuleTransformation 129  The first step of the pipeline corresponds to the input of multiple high-dimensional omic datasets from the same individuals as well as subject phenotypes (two or more groups). The next step consists of data normalization (specific to each dataset), and pre-filtering of irrelevant features. Prior to data integration, exploratory data analysis can be performed using existing mixOmics methodologies such as principal component analysis (PCA). DIABLO is then used to construct a multi-omic biomarker panel consisting of a subset of variables from each omic dataset. For a new subject, predictions are made by each omic panel within the multi-omic biomarker panel. These predictions are combined by averaging all the predictions from each omic panel to determine the consensus prediction for a new subject. The performance of this panel was estimated using a M-fold cross-validation that is repeated k times (a leave-one-out cross-validation was performed for all analyses in this chapter, where M = N, number of samples) and the area under the receiver operating curve (AUC) was determined from the average predictions. Since outputs of biomarker algorithms can be challenging to interpret, numerous visualizations were implemented to aid in the interpretation of the multi-omic biomarker panel: 1) sample plots depicting subject clustering based on phenotypic labels, 2) circos plots showing the relationship between variables, and 3) heatmaps depicting the expression or the correlation between variables.  5.3.3! Simulation study Prior to applying DIABLO to the asthma datasets, a simulation study was performed to study the relationship between the design matrix and the types of variables selected by the DIABLO classifier and its corresponding classification error rate. Two design matrices were tested; the full  130 design (all datasets were connected) and the null design (where no datasets were connected). Three datasets X, Y, and Z were generated with equal numbers of observations (n=100) and 150 variables for two groups of 50 observations each. 100 variables were irrelevant variables (not correlated between datasets, and not discriminatory), whereas the 50 variables were either 1) correlated across all three datasets but not discriminatory between groups (CorNonDis), or 2) correlated and discriminatory (CorDis) or 3) not correlated but discriminatory (NonCorDis) (see Appendix D.4 for details). Separate scenarios of three datasets were constructed for each variable type of size 100 x 150, in order to assess how DIABLO functions under these scenarios. The fold-change between the two groups was varied from 0 to 1.5, and varying levels of noise were added. Furthermore, the strength of correlation was varied, as shown in Figure 5.4 which depicts three covariance matrices for low (!=0.75), medium (!=0.91) and highly (!=0.98) correlated variables.  Low Medium High    Figure 5.4 Covariance matrices varying the strength of correlation between variables. The strength of the correlation between variables was varied to determine its effects on the number of correctly identified variables by the DIABLO classifier with respect to the design matrix specification.   ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● −1−0.8−0.6−0.4−0.200.20.40.60.811 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45123456789101112131415161718192021222324252627282930313233343536373839404142434445●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● −1−0.8−0.6−0.4−0.200.20.40.60.811 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45123456789101112131415161718192021222324252627282930313233343536373839404142434445●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● −1−0.8−0.6−0.4−0.200.20.40.60.811 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45123456789101112131415161718192021222324252627282930313233343536373839404142434445 131 Simulated datasets were generated 20 times for each type of variables and DIABLO was applied with the full and null design selecting 50 variables and retaining one component. As expected under the full design (all datasets are connected, Appendix D.3), DIABLO correctly identified a greater proportion of CorNonDis variables compared to the null design (no connected datasets) (see green solid line vs. green dashed line in Figure 5.5) and this difference increased as the strength of correlation between the variables was increased. Similarly, DIABLO selected a greater proportion of CorDis variables under the full design compared to the null design, however this difference decreased as the fold-change increased (red solid lines vs. red dashed lines). Interestingly, no difference was found between the full and null designs for NonCorDis variables (blue solid lines vs. blue dashed lines).  no noise low noise medium noise high noise●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● ● ●●● ●●●●●●●●●●●● ● ● ● ●●●●●●●●●●●● ● ● ● ●●●●●●●●●●●● ● ● ● ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● ●●● ●●●●●●●●●●●●●● ●●●●●●●●● ● ● ●●●●●●●●●●●● ● ● ● ●●●●●●●●●●●● ● ● ● ●●●●●●●●●●●●●●●●●●●●●●●●●●● ●●●●●●●●● ● ● ●●●● ●●●●●●●● ●●●●●●●● ● ●●●●●● ●●●●●●●●● ● ●●●●●●● ● ●●●●●●●●●●●●●● ● ● ● ●●●●● ●●●●●●●●●●● ●●● ●● ● ●●●●●●●●●●● ● ● ●●● ● ● ●●●●●● ●●●●●●● ● ● ●●●●●● ●● ●●●●●● ●●●● ●●●●●● ● ● ●● ●●●●●●●●● ●●● ●●●●●● ● ●0.250.500.751.000.250.500.751.000.250.500.751.000.250.500.751.000.250.500.751.0000.3750.751.1251.50.75 0.80 0.85 0.90 0.95 0.75 0.80 0.85 0.90 0.95 0.75 0.80 0.85 0.90 0.95 0.75 0.80 0.85 0.90 0.95strength of correlationAverage proportion of correctly identified variables (20 permuations)varType●●●CorDisCorNonDisNonCorDisDesignFullNull0 0.5 1 1.5●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● ●●●●●●●●●●●●●●●● ● ●●● ●●● ●●●●●●●●●●●●● ● ● ●● ●● ● ●●●●●●●●●●●● ● ● ● ●●●●●●●●●●●●●●●●●●●●●●●●● ● ●●●●●●●●●●●● ●●●●●●●●●●● ●●●●●●●●●●●●●●●●●●● ●● ●●●●● ● ●●●●●●●●●●●●●●● ● ● ● ●●●● ●● ●●●●●●●●●●●●● ● ● ● ●●●●●●●●●●●●●●●●●●●●●● ●●●●●●●●●●●●● ●●●●●●●●●● ●●●●●●●●●●●●●●●●●● ●●●●●●●●●●●●●● ●●● ●●●●●●●●●● ●● ●● ●●●●● ● ●●●●●●●●●●●● ● ●● ● ●● ●●●●●●●●●●●●●●●●●●●●●● ●● ●●●●●●●●●●●●●●●●●●●●●●●●●● ●●●●●●●●●●●●●●● ●●●●●●●●●●●●●●●● ●●●●●●●●● ●●●●●●●● ●●●● ●●●●●●●●●●●● ●● ●● ●●●●0.250.500.751.000.250.500.751.000.250.500.751.000.250.500.751.000.250.500.751.0000.3750.751.1251.50.75 0.80 0.85 0.90 0.95 0.75 0.80 0.85 0.90 0.95 0.75 0.80 0.85 0.90 0.95 0.75 0.80 0.85 0.90 0.95strength of correlationProportion of correctly identified variablesvarType●●●CorDisCorNonDisNonCorDisDesignFullNullFold-change  132 Figure 5.5 Relationship between the design and variables selected in the DIABLO model. Under the full design, DIABLO correctly identified a greater proportion of CorNonDis variables (green) compared to the null design and this difference increased as the strength of correlation was increased (green solid line vs. green dashed line). The CorDis variables (red) were selected at a greater proportion under the full design compared to the null design, however this difference decreased as the fold-change increased (red solid line vs. red dashed line). No difference was identified for NonCorDis variables (blue) between the full and null design for any fold-change, correlation or noise threshold.   No significant difference was observed between the full and null design with respect to the error rate (Figure 5.6) for any type of variable. The error rate was lower when the datasets contained NonCorDis compared to the CorDis variables and this difference increased as the correlation of the CorDis variables increased, but this difference diminished as the fold-change increased (red vs. blue lines). The green lines depict DIABLO applied to datasets with CorNonDis and irrelevant variables and as expected, the error rate averages around 50% (random prediction of the DIABLO model). Therefore, although the design matrix may not affect the error rate of the DIABLO classifier, the design determines the types of variables that are retained by the DIABLO model.   133   Figure 5.6 Error rates of various DIABLO models under different designs, types of variables, strength of correlation, fold-change and noise. For a fold-change of zero, all models had an average error rate of 0.50 (random prediction). However, as the fold-change increased, the DIABLO model with NonCorDis variables outperformed the DIABLO model with CorDis variables regardless of the design, however this difference diminished with increased fold-change. The DIABLO model applied to datasets that only contained CorNonDis variables and irrelevant variables had an error rate around 0.50, since these variables were not discriminatory.  5.3.4! Study 1: Changes in molecular pathways in response to allergen challenge All asthmatic individuals undergoing allergen inhalation challenge elicited the early asthmatic response (Figure 5.7A). In order to obtain a holistic view of the underlying molecular interactions that take place in peripheral blood, DIABLO was used to integrate cell-counts, gene no noise low noise medium noise high noise●● ●● ●●●● ●●●●●●●●●●● ●●●● ●●●●●●●● ●●● ●●●●●● ●●●●●● ●●●●●●●●● ●●●● ●●●● ● ● ● ●●●●●●●●●●● ●●●● ● ● ●●● ●● ●●●● ●●● ● ●●●● ●●●●●●● ●●●●●●●●● ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● ●●● ●●●● ● ● ● ●●●●●●● ●●●●●●● ●●● ● ● ●●●●●●●●●● ● ●●●● ●●●●●● ● ●●● ●●●●●●●●● ●● ● ●●●●● ●●●●●●●●●●●●●● ● ●●● ●●●●●●●●●●●●●●●●●● ●●●●●●●● ●●●●● ● ● ● ●●●● ●●●● ●●● ●● ●●● ● ●●●●●●●●●●●●● ●●● ●●●●● ●●●●● ●● ●● ●●●●●●●●●●● ●●●●●●●● ●●●●●●●● ●●●●●●●●●●● ●●●●● ●● ●●● ●●●●●●●●●●●● ●●●● ●● ●● ●●●● ●● ●●●0.00.20.40.60.00.20.40.60.00.20.40.60.00.20.40.60.00.20.40.600.3750.751.1251.50.75 0.80 0.85 0.90 0.95 0.75 0.80 0.85 0.90 0.95 0.75 0.80 0.85 0.90 0.95 0.75 0.80 0.85 0.90 0.95strength of correlationError rate (10x−5−fold cross−validation)varType●●●CorDisCorNonDisNonCorDisDesignFullNull0 0.5 1 1.5●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● ●●●●●●●●●●●●●●●● ● ●●● ●●● ●●●●●●●●●●●● ● ● ●● ●● ● ●●●●●●●●●●● ● ● ●●●●●●●●●●●●●●●●●●●●●●●●● ● ●●●●●●●●●●●● ●●●●●●●●●●● ●●●●●●●●●●●●●●●●●●● ●● ●●●●● ● ●●●●●●●●●●●●●●● ● ● ● ●●●●● ●●●●●●●●●●●● ● ● ● ●●●●●●●●●●●●●●●●●●●●●● ●●●●●●●●●●●●● ●●●●●●●●●● ●●●●●●●●●●●●●●●●●● ●●●●●●●●●●●●●● ●●● ●●●●●●●●●● ●● ●● ●●●●● ● ●●●●●●●●●●●● ● ●● ● ●● ●●●●●●●●●●●●●●●●●●●●●● ●● ●●●●●●●●●●●●●●●●●●●●●●●●●● ●●●●●●●●●●●●●●● ●●●●●●●●●●●●●●●● ●●●●●●●●● ●●●●●●●● ●●●● ●●●●●●●●●●●● ●● ●● ●●●●0.250.500.751.000.250.500.751.000.250.500.751.000.250.500.751.000.25.500.751.0000.3750.751.1251.50.75 0.80 0.85 0.90 0.95 0.75 0.80 0.85 0.90 0.95 0.75 0.80 0.85 0.90 0.95 0.75 0.80 0.85 0.90 0.95strength of correlationProportion of correctly identified variablesvarType●●●CorDisCorNonDisNonCorDisDesignFullNullFold-change  134 expression and metabolite abundances using the datasets from Chapter 3. A module based approach was used to transform both the gene expression (microarray) and metabolite (Metabolon) datasets into pathway datasets, such that instead of gene/metabolite activity, each variable represented the pathway activity level in each sample. The mRNA dataset was transformed into a Kyoto Encyclopedia of Genes and Genomes (KEGG) dataset by extracting all genes for a given pathway from the mRNA dataset and then applying PCA and retaining the first principal component (these new variables can be interpreted as the expression (activity) of a particular pathway). Similarly, the metabolite dataset was transformed into a metabolite pathway dataset. Metabolite modules were highly (Pearson correlation > 0.8) correlated with cell-counts and gene modules (Figure 5.7B).    135  Figure 5.7 Systems approach to molecular changes in blood after allergen inhalation challenge. A. FEV1 response profiles 0-2 hours after allergen inhalation. B. Design matrix used in the DIABLO model, depicting the pair-wise correlation between datasets. C. Path diagram of the connection between datasets, all predicting the time point variable. The mRNA and metabolite datasets were transformed into module datasets. D. Sample plots depicting the clustering of subjects based on the first component of each dataset from the DIABLO model. E. Correlation between variables selected in the DIABLO model.  Figure 5.7C shows the connectivity between datasets used in the DIABLO model to identify correlated sets of cells, gene and metabolite modules that were altered after allergen inhalation challenge. For each dataset, the within-sample variation matrix was extracted (analogous to taking the differences between paired samples) using a previously described approach236. The −1−0.8−0.6−0.4−0.200.20.40.60.81CellsgeneModsmetModsCellsgeneModsmetMods10.520.8510.82 1CellsgeneModsmetMods●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●−50−40−30−20−100100 25 50 75 100 125Time (minutes)Percent drop in  FEV1A. B. C.E.D.mRNA Gene(modulesMetabolite(modulesCellsPrevsPostMetabolitesConnect+datasets−1 0 0.5 1Color keyValine, leucine and isoleucine biosynthesisThyroid cancerValine, leucine and isoleucine metabolismTryptophan metabolismFood component/PlantVitamin B6 metabolismLysolipidButanoate metabolismFatty acid metabolismGraft−versus−host diseaseAsthmaStaphylococcus aureus infectionAutoimmune thyroid diseaseAllograft rejectionRelative.EosinophilsDipeptideRelative.BasophilsAlanine and aspartate metabolismgamma−glutamylTryptophan metabolismDrug metabolism − other enzymesGlycerolipid metabolismValine, leucine and isoleucine biosynthesisThyroid cancerValine, leucine and isoleucine metabolismTryptophan metabolismFood component/PlantVitamin B6 metabolismLysolipidButanoate metabolismFatty acid metabolismGraft−versus−host diseaseAsthmaStaphylococcus aureus infectionAutoimmune thyroid diseaseAllograft rejectionRelative.EosinophilsDipeptideRelative.BasophilsAlanine and aspartate metabolismgamma−glutamylTryptophan metabolismDrug metabolism − other enzymesGlycerolipid metabolism●●●CellsGeneModulesMetModulesIndexCells●●●●●●●●●●●●●●●●●●●●●●●●●●●●X.labelY.label−2 −1 0 1 2●●●●●●●●● ●●●●●●●●●●●●●●●●●●●X.labelY.label−2 −1 0 1 2−4−2024prepostIndex−0.55Index1 geneMods●●●●●●●●●●●●●●●●●●●●●●●●●●●●X.labelY.label−4−2024prepost−0.62 1 0.72 1 metModspre post03.5710.514 136 DIABLO model consisted of 2 cell-types, 10 gene and metabolite modules across two components with AUC=99% (leave-one-out cross-validation). Figure 5.7D shows a clear separation between pre- and post-challenge samples based on the variables selected by DIABLO (Figure 5.7E). Interestingly some gene-metabolite modules with the same ontologies such as Valine, leucine and isoleucine metabolism (biosynthesis) and Tryptophan metabolism were selected in the DIABLO model. Other selected modules such as Glycerolipid metabolism (gene module), Lysolipid and Fatty acid metabolism (metabolite modules) were related to lipid metabolism. An important gene-module that was selected in the DIABLO model was the Asthma  pathway (Figure 5.8). Although Figure 5.8 indicates increased Interleukin (IL)-4 and Fc"RI expression and decreased expression of IL-13, and Major Histocompatibility Complex class II (MHCII) these changes were not significant at BH-FDR < 5%.   Figure 5.8 Regulation of asthma genes in response to allergen inhalation challenge in asthmatics. Red denotes up-regulation of gene expression two hours after challenge, whereas green denotes down-regulation after challenge. Of note is that although the Asthma pathway was selected using DIABLO, at the single gene level, no gene expression differences between pre and post challenge were identified at BH-FDR=5%.  137  The 14 asthmatic subjects consisted of 8 isolated early responders (ERs) and 6 dual responders (DRs) (Table 3.1). Differential patterns were identified when comparing ERs and DRs (Figure 5.9). ERs demonstrated an eosinophilic response through increased expression of Eotaxin, IL-3 and Eosinophil peroxidase (EPO). Interesting the pro-inflammatory tumor necrosis factor alpha (TNF-#) decreased in expression where the anti-inflammatory cytokine (IL-10) increased in expression. A heightened inflammatory response was observed in DRs, through up-regulation of Fc"RI, and pro-inflammatory genes such as IL-4, and IL-5 as well as decreased expression of IL-10. The expression of IL-13, another pro-inflammatory cytokine decreased after allergen challenge in DRs in the blood.   138 Early Responders  Dual Responders   139 Figure 5.9 Regulation of asthma genes in response to allergen inhalation challenge in asthmatics in ERs and DRs. A more pronounced inflammatory response was observed in DRs (lower panel) compared to ERs (upper panel). Red denotes up-regulation of gene expression two hours after challenge, whereas green denotes down-regulation after challenge.  Comparing the change in gene expression in ERs with the change in gene expression in DRs identified no differentially expressed genes except for IL-10 (BH-FDR <5%). IL-10 gene expression (microarray) increased in ERs but decreased in DRs after allergen inhalation challenge (Figure 5.10 left panel) comparing 8 ERs with 6 DRs. IL-10 counts from the RNA-Seq (Ensembl) dataset from Chapter 4 were used to validate this signal in an independent cohort (Table 4.1). However, IL-10 expression (RNA-Seq) was not differentially expressed between ERs and DRs comparing 15 ERs and 21 DRs. The expression of IL-10 in the RNA-Seq data was very low for most samples, with counts below zero after normalization (Figure 5.10 right panel).    140  Figure 5.10 Differential expression of IL-10 between ERs and DRs in response to allergen challenge. Comparing the change (post minus pre) in gene expression in ERs with the change in gene expression in DRs of all genes in the Asthma pathway, only IL-10 was statistically significant. IL-10 gene expression decreased in DRs post-challenge whereas its expression increased in ERs post-challenge based on microarray data (left). This difference was not statistically significant using RNA-Seq data from an independent cohort (right).  5.3.5! Study 2: Multi-omic biomarker panel of the late phase asthmatic response The module approach was useful in improving biological interpretability of the underlying biological interactions and may be predictive in discriminating clinical phenotypes. However, since each module is a collection of genes or metabolites, and many modules are part of a multi-omic biomarker panel, the number of single molecule candidates can far exceed the cost to develop and implement these biomarker panels. Therefore, a multi-omic biomarker panel that consists of only a few molecules from each omic dataset may be a more feasible option for clinical implementation. ●●3.503.754.00Pre PostTimelog2 expressionResponseDRERIL−10 (Microarray)●−4−3−2−101Pre PostTimelog2 cpm ResponseDRERIL−10 (RNA−Seq) 141 789 genes corresponding to 84 cell-types were extracted from the unfiltered Ensembl RNA-Seq dataset (from Chapter 4) and used to construct surrogate variables representative of cell-type frequencies (see details in Section 5.2.2). The filtered Ensembl dataset consisted of 7,518 transcripts with no overlap with the 789 cell marker genes. The results of Chapter 3 implicated a strong role of lipid metabolism between early and dual responders, therefore a targeted metabolomics kit was used to profile some acylcarnitines and amino acids but mostly lipid metabolites such as glycerophospholipids and sphinolipids. Three datasets of sizes 32 (subjects) x 84 cell-types, 32 x 7,518 transcripts and 32 x 163 metabolites were used to discriminate ERs from DRs using DIABLO. In order to identify the most discriminatory molecules from each dataset, a null design matrix (where no datasets were connected) was used to construct the DIABLO classification model, retaining 4 cells, 20 mRNA transcripts and 10 metabolites across two components (Figure 5.11A). The sample clustering based on the multi-omic biomarker panel showed a modest separation using the first component from each omic dataset (Figure 5.11B). Highly expressed genes in DRs compared to ERs (orange vs. blue line) were positively correlated with variables from other datasets whereas genes with lower abundance in DRs relative to ERs were negatively correlated (Figure 5.11C). The AUC based on a leave-one-out cross-validation was 68.6% for the DIABLO classifier (Figure 5.11D), although the predictions of the single omic biomarker panels ranged in AUCs from 59.2% (genes) to 88.6% (metabolites).    142  Figure 5.11 Multi-omic biomarker panel predicts ERs and DRs prior to challenge. A. Path diagram depicting the design between datasets. B. Sample clustering of ERs and DRs based on the fist component of the DIABLO model. C. A circos plot showing the association (score greater than 0.5) between variables of each omic biomarkers. The circos plot consists of an ideogram consisting of variables color-coded to the different omic datasets. A surrounding line graph depicts the expression of each variable in ERs and DRs, respectively. For a correlation cut-off of 0.5, the links corresponding to positive (red) and negative (blue) associations between variables are plotted. D. AUCs based on a leave-one-out cross-validation using prediction made by each datasets from the DIABLO model and the average prediction of all datasets.  The multi-omic classifier include cell-types such as Plasma cells, CD8 T cells, HSC-3 cell line and CD4 T cells. The 20 selected genes were CDK1, TMEM101, DESI1, MRPL32, TREML2, SF3B1, IVNS1ABP, NSL1, SLC17A5, RBM39, KDELR2, HLX, SLC15A4, NXPE3, HK1, SCIMP,  143 SLC3A2, CEP97, CHP1, and ZKSCAN4. Selected metabolites included acylcarnitines (C5-DC (C6-OH) – Glutarylcarnitine, C10:2 – Decadienylcarnitne, C18:1 – Octadecenoylcarnitine), amino acids (tyrosine, leucine/isoleucine) and glycerophospholipids (lysoPC a (lysoPhosphatidylcholine acyl) C14:0, lysoPC a C28:0, Phosphatidylcholine acyl-alkyl (PC ae) C30:0, PC ae C40:6, PC ae C44:3). Pathway over-representation analysis using InnateDB was performed, but instead of using the genes from the multi-omic biomarker panel, a longer gene list consisting of genes from all multi-omic biomarker panels from the folds of the leave-one-out cross-validation were used (Figure 5.12). The top ranked pathways included Synthesis of Leukotrienes (LT) and Eoxins (EX), Prostaglandin Leukotriene metabolism and arachidonic acid metabolism which included genes such as arachidonate 5-lipoxygenase (ALOX5), cytochrome P450, family 4, subfamily F, polypeptide 3 (CYPAF3) and leukotriene A4 hydrolase (LTA4H) (BH-FDR <10%). VEGFR2 mediated cell proliferation was also highly ranked (BH-FDR<10%) and included genes such as mitogen-activated protein kinase 1 (MAPK1), protein kinase C, beta (PRKCB) and tyrosine kinase 2 (TYK2). Pathways related to amino acid metabolism included solute leukocyte carrier SLC-mediated transmembrane transport and transport of inorganic cations/anions and amino acids/oligopeptides and consisted of panel genes such as hexokinase 1 (HK1), solute carrier family 15, member 4 (SLC15A4), solute carrier family 17 (anion/sugar transporter), member 5 (SLC17A5) and solute carrier family 3 (activators of dibasic and neutral amino acid transport), member 2 (SLC3A2).  144  Figure 5.12 Pathway over-representation analysis using InnateDB. Enriched pathways in the list of genes comprised of genes from multi-omic biomarker panels in the folds of the leave-one-out cross-validation. Many lipid synthesis and metabolism pathways were identified as discriminatory between ERs and DRs.  5.3.6! Multi-omic biomarker panel in response to allergen inhalation challenge The multi-omic biomarker panel from Figure 5.11 had a classification performance based cross-validation of AUC=68.6% using the average predictions from all the three datasets. The panel of genes had the weakest performance at AUC=59.2%, whereas the metabolite had the highest Significanc  threshold: 1.3-Log10 PvalueAllSynthesis of Leukotrienes (LT) and Eoxins (EX)Prostaglandin Leukotriene metabolismVEGFR2 mediated cell proliferationERK2 activationStat3 signaling pathwayIL11ERK activationTransport of inorganic cations/anions and amino acids/oligopeptidesArachidonic acid metabolismActivation of BAD and translocation to mitochondriaSLC-mediated transmembrane transportRespiratory electron transport, ATP synthesis by chemiosmotic coupling, and heat production by uncoupling proteins.Cadmium induces dna synthesis and proliferation in macrophagesRAF/MAP kinase cascadeOocyte meiosisAspirin blocks signaling pathway involved in platelet activation01234 145 performance at AUC=88.6%. The post-challenge expression of multi-omic biomarker panel candidates was extracted and normalized to the pre-challenge expression level. Using leave-one-out cross-validation the classification performance was determined for the multi-omic biomarker panel. The AUC of the multi-omic biomarker panel at post-challenge remained similar to that of the pre-challenge combined performance (Figure 5.13). The predictions from the subset of genes in the multi-omic biomarker panel resulted in the highest performance (AUC=77.7%) post-challenge.  Figure 5.13 AUC of the multi-omic biomarker panel post-challenge using a leave-one-out cross-validation. The AUC of the multi-omic biomarker panel at post-challenge was similar to that at pre-challenge at 67.8%. Note the differences between pre- and post-challenge levels for all datasets was determined and the classification performance was determined using a leave-one-out cross-validation such that same multi-omic biomarker panel was evaluated in each fold.  5.4! Discussion These results are consistent with the findings of many studies that have implicated the various cell-types, genes and metabolites and their role in the pathogenesis of asthma and their response to different experimental conditions. More importantly this work identified novel pathways and molecules that were highly predictive of the late phase asthmatic response. ● ● ● ● ●●●●●●● ●● ● ●● ●● ● ● ● ●●● ● ● ●●● ● ●●●0 20 40 60 80 100020406080100Cells100−SpecificitySensitivityAUC = 48.2%●●● ●●●●●● ● ● ● ●●●● ●●● ●●● ● ● ● ●● ● ● ● ● ● ●0 20 40 60 80 100020406080100Genes100−SpecificitySensitivityAUC = 77.7%● ●● ●●●●●● ● ●● ●● ●● ● ●● ● ●● ●●● ●●● ● ● ● ● ●0 20 40 60 80 100020406080100Metabolites100−SpecificitySensitivityAUC = 64.7%●● ●●●●● ● ● ● ●●●● ● ● ● ●●●●●●●● ● ● ● ● ● ● ● ●0 20 40 60 80 100020406080100Combined100−SpecificitySensitivityAUC = 67.8% 146  DIABLO was developed for the purposes of data integration and discriminant analysis using multiple high-dimensional omic datasets. DIABLO is an iterative algorithm that aims to maximize the correlation between different high-dimensional datasets with the goal of selecting a subset of highly correlated and discriminatory variables. DIABLO was used to provide a systems view of cellular and molecular pathways in the blood of asthmatics in response to allergen inhalation challenge and between ERs and DRs prior to allergen inhalation challenge. Levels of eosinophils and basophils237, two hallmark cells of allergic asthma were significantly altered two hours after allergen challenge. Gauvreau et al.29, showed an increase in sputum eosinophils and basophils 7 and 24 hours after allergen inhalation challenge. In addition, higher levels of basophils were observed in DRs compared to ERs and were correlated with airway hyperresponsiveness (% drop in FEV1 between 3-7 hours after challenge). Basophils and eosinophils are derived from CD34+ progenitor cells in the bone marrow and recruited to sites of inflammation. Eosinophils express a range of cytokines and chemokines such as TNF-!, IL-3, IL-4, IL-10 and IL-13237 all of which were implicated in the Asthma KEGG pathway selected using DIABLO. With the exception of IL-10, basophils can also produce these proteins as well as induce the production of IL-10 in T cells237–240. IL-10 expression was seen to increase in ERs and decrease in DRs two hours after allergen challenge in the microarray data, but not using the RNA-Seq data of an independent cohort. However, it is important to note the low coverage of IL-10 transcripts in this dataset (normalized counts less than zero), and a higher depth of coverage should be used to obtain reliable results. Another study has shown similar patterns but with IL-10-producing CD4+ T cells from PBMCs, 24 hours after challenge30. DIABLO also selected the valine, leucine and isoleucine (branched-chain amino acids, BCAAs) biosynthesis gene module and the valine, leucine and isoleucine metabolism metabolite module, where the  147 activity of both significantly increased post-challenge. A related gene module, valine, leucine and isoleucine degradation pathway was not selected by DIABLO, and was not significant (p=0.15) and decreased in activity post challenge. Interestingly, xleucine (valine and isoleucine) was also identified in the predictive multi-omic biomarker panel of the late phase asthmatic response. The transport of these amino acids in and out of cells may be mediated by solute carrier proteins241, also enriched in the multi-omic biomarker panel. This amino acid metabolism pathway has only been recently linked with asthma endotypes where higher levels of plasma BCAAs (valine, leucine and isoleucine) were observed in asthmatics with high levels of fractional nitric oxide levels (FENO) compared to asthmatics with low FENO242. FENO is associated with airway inflammation (sputum eosinophils), higher serum IgE and higher blood eosinophils243 and can also predict response to therapy163. However its role as an asthma biomarker remains unclear with conflicting evidence and thus should be assessed in combination with other mediators58. The mechanistic relationship between BCAAs and allergic responses and their role as potential predictor of allergic inflammation needs to be further investigated.  A recurring theme in this thesis has been the role of lipid mediators in the development of asthmatic responses. Lipid mediators play a central role in the allergic response, with the major cell-types such as mast cells, basophils and eosinophils all capable of producing de novo lipid metabolites such as prostaglandins (e.g. PGD2) and leukotrienes (e.g. LTC4)237,244 from arachidonic acid, a polyunsaturated fatty acid bound to membrane phospholipids. The multi-omic biomarker panel consisting of cells, gene transcripts and metabolites that was predictive of the late phase asthmatic response also implicated the role of T-cells, which is supported by other studies demonstrating that LTB4 and PGD2 can recruit T cells to the airways245. Glycerophospholipids that were part of the multi-omic biomarker panel were much stronger  148 predictors of the late phase asthmatic response compared to the selected cell-types and gene transcripts. Upon allergen inhalation, phosphatidylcholines may be metabolized by phospholipase A2 (PLA2) into mediators such as prostaglandins and leukotrienes leading to smooth muscle contraction and induction of leukocyte chemotaxis (e.g. of T cells)245. Since increased levels of PLA2 are released into the airways during asthma exacerbations246, higher baseline levels of phosphatidylcholines in dual responders may increase their susceptibility to the late phase asthmatic response. Using the same lipid profiling kit (Biocrates Life Sciences), previous studies have implicated the role of phosphatidylcholines in asthma and atopic dermatitis247,248. In addition, increased concentrations of phosphatidylcholines have been associated with risk alleles in the asthma susceptibility locus 17q21247, but whether these risk alleles are associated with the late response remains to be investigated. Current lipid biomarkers of asthma include arachidonic acid metabolites such as leukotriene E4 (LTE4) and lipoxin A4 (LXA4). Urinary LTE4 has been shown to be increased in allergic asthmatics and is associated with asthma exacerbations64. The ratio of urinary LTE4 to FENO has been shown to better predict the response to leukotriene receptor antagonists compared to inhaled corticosteroids249. LXA4 is an anti-inflammatory mediator that is decreased in severe asthma, potentially related to systemic corticosteroid treatment250. The results described in this chapter further suggest that plasma may be a useful medium that holds promise in identifying novel diagnostic biomarkers of asthma.  Although the performance of individual omic panels in the multi-omic biomarker panel varied considerably in predicting the late response, two hours post challenge compared to pre-challenge, the combined performance was similar (AUC = 68%). This suggests using combinations of different biological markers may compensate for fluctuations at the individual omic level, although multiple time-series blood draws will be required to assess this hypothesis.  149 Interestingly, the gene transcript biomarkers increased in their predictive ability after challenge, which may be indicative of a mechanistic link with activation during the late phase response. Future studies will be required to validate the multi-omic biomarker panel identified in this chapter.  5.5! Limitations The integrative classification method developed in this section performed well in identifying variables across different omic datasets that may be participating in common underlying biological processes. However, finer tuning of this algorithm, such as identifying the optimal number of variables and components to select for each omic datasets, needs further development. Furthermore, predictions from each dataset were averaged, resulting in a compromise between the weakest and strongest omic panels. Other methods of combining multiple predictions such as majority voting rules251 could be considered in future work. However, the post-challenge performance of the multi-omic biomarker panel was similar to that at pre-challenge which may indicate the increased utility of averaging multiple biomarker panels. On the other hand, the post-challenge performance may be biased as the expression levels were normalized using pre-challenge expression (post minus pre) prior to performing the leave-one-out cross-validation.  Each omic dataset was generated from different biological materials using different technologies, each with its own set of limitations. The cell-type dataset was obtained using statistical deconvolution of gene expression data since cell-type frequencies were unavailable. The group (phenotype) effect was removed through statistical means, although this may not have entirely removed the bias. Furthermore, the depth of coverage of the RNA-Seq data may affect accurate quantification of low abundant cell-specific marker genes. The present study did not  150 control for sleep deprivation, diet and exercise, all of which can affect metabolite levels252,253. However, all challenges were conducted in the morning at similar times to minimize differences due to circadian fluctuations of metabolite levels. Despite these limitations, the multi-omic biomarker panel showed good classification performance, but these results need to be validated in an independent external cohort.  5.6! Conclusion The results of this chapter show that combination of molecules across different biological layers may accurately capture dysregulated disease processes and may even be used as diagnostic tools to identify individuals at risk of developing the late asthmatic response. A classification approach that incorporates the correlation structure between biological layers should enable improved biomarkers for sub-phenotyping, predicting asthma exacerbations and even response to therapy.     151 Chapter 6:!Conclusions & future directions 6.1! Conclusions This dissertation focused on delineating molecular differences in peripheral blood between allergen-induced airway responses. The use of unbiased high throughput technologies coupled with data-driven computational methods enabled the identification of cell-types and many molecules from different compartments of peripheral blood, taking part in various molecular processes that lead to the clinical manifestations called the early and dual asthmatic responses. Therefore, I conclude that significant molecular changes in the blood exist between isolated early responders (ERs) and dual responders (DRs) which may be due to differences in the underlying molecular mechanisms. Although no mechanistic studies were performed in this thesis work and all conclusions are based on associations, there is an overwhelming amount of data to suggest distinct molecular phenotypes of ERs and DRs that are different at baseline (prior to allergen inhalation) and respond differentially to allergen inhalation challenge. Furthermore, biomarker panels that may represent these molecular phenotypes can be used to identify asthmatic individuals that may possess characteristics of DRs. These features are characteristic of pro-inflammatory and dampened anti-inflammatory phenotypes that result in chronic inflammation, airway hyperresponsiveness, and airway remodeling.  6.2! Review of aims In Chapter 2, the utility of the allergen inhalation challenge model in identifying molecular differences between pre and post challenge254,255 and between ERs and DRs256 was demonstrated. These molecular changes were characteristics of lipid metabolism and leukocyte migration, supporting evidence from previous studies that have implicated their roles in airway  152 response to allergen challenge. MicroRNA (miR)-192 was identified as under-expressed in asthmatics compared to healthy controls and was down-regulated post-challenge. This was a novel finding as no previous study has associated miR-192 with asthma pathogenesis to the best of my knowledge. Plasma fibronectin was found to be elevated in DRs compared to ERs. These results clearly demonstrate the usefulness of the allergen inhalation challenge in identifying molecular changes in response to allergen challenge and between asthma endotypes.  In Chapter 3, gene-metabolite expression changes in response to challenge that were different between early and dual responders were enriched in arachidonic acid metabolism257. Differences in other mediators such as cortisol and bradykinin were also identified. Further investigations of the other identified targets may provide insight into novel molecular mechanisms. The Th17/Treg ratio representing a pro- and anti-inflammatory axis was significantly associated with the late phase asthmatic response175. Furthermore, this ratio was correlated with gene expression profiles of genes belonging to the leukocyte receptor complex, which are expressed by NK cells.  In Chapter 4, differential gene expression analyses identified significantly stronger changes between ERs and DRs prior to allergen inhalation challenge compared to post-challenge. Biomarker panels that could identify asthmatic individuals who were likely to develop the late phase response were developed. These panels were recalibrated using the NanoString platform and successfully validated in an independent external cohort (AUC~70%). These may be used as blood tests to identify at risk/susceptible asthmatic individuals with indications of chronic airway inflammation.  In Chapter 5, an integrative classification method was developed and applied to datasets from Chapter 3 in order to obtain a holistic view of the molecular changes in response to allergen  153 challenge. This algorithm was also used to develop a multi-omic biomarker panel that was predictive of the late phase asthmatic response at pre allergen challenge and performed similarly post allergen challenge (AUC~68%). These panels may indicate common underlying pathways that span different biological layers and are more representative of asthma pathobiology.  6.3! Strengths and limitations Asthmatic patients with severe uncontrolled asthma make up a fraction of the asthma population yet incur the majority of the costs258. Although this is an unmet clinical need, the majority of asthma patients have mild to moderate asthma which is often underdiagnosed and undertreated259. Airway inflammation and epithelium damage have been observed in subjects with mild disease260,261. Therefore, treating mild asthma may slow down the progression of chronic inflammation, airway hyperresponsiveness and airway remodeling.   The use of blood to study respiratory diseases has been questioned as being too systemic in nature and not reflecting the inflammatory processes in the lung. However, blood is the medium of choice for clinical tests since its collection is minimally invasive, cost-effective, and it is relatively easy to obtain. The work in this thesis demonstrates that profiles in the blood can be associated with allergic asthma which may be due to the crosstalk between the lung and the bone marrow-derived immune cells. Although the allergen inhalation challenge can be used to study the physiological, cellular, and molecular changes of airway responses, accurate phenotypic characterization of subjects into ERs and DRs is still difficult. Spirometry is dependent on patient cooperation and effort and must be repeated several times7. Allergen specificity for the late phase asthmatic response can dictate the type of response an individual elicits156. Furthermore, the duration of  154 direct observation of a subject undergoing the allergen inhalation challenge is only 7-8 hours, such that subjects who develop the late response after this time point may be misclassified21. Since these subjects are recruited for clinical trials where preference is given to DRs, individuals screened as ERs are not rescreened in the future, whereas DRs can be rescreened for different trials thereby establishing the true phenotype of these individuals. The group of subjects used in this thesis represent a homogenous group of non-smokers with mild, stable, atopic asthma, free of other lung diseases and are not using regular anti-inflammatory treatments. This helps decrease the bias introduced by confounding variables and improves the chances of detecting true differences between responder groups. The results of this study may directly impact the subject recruitment for clinical trials aimed at testing new treatments for asthma in their ability to attenuate the late phase asthmatic response.  6.4! Future directions This project has contributed to accelerating Canadian innovation through the creation of novel biomarker panels to predict late phase asthmatic responses. These biomarker panels demonstrated good performance for predicting the allergen-induced late phase asthmatic response. Their utility may be further evaluated in screening for asthmatic late responders to allow more effective recruitment into clinical trials for new asthma drugs directed towards attenuating the late response. The late response has been linked to chronic asthma inflammatory phenotypes, and these biomarker panels may have additional clinical utility in the Canadian health system and internationally. In addition, this biomarker blood test may be useful for other allergic conditions, including allergic rhinitis, anaphylaxis, and skin reactions, where late-phase responses have also been documented and are of clinical significance.  155 References 1.  World Health Organization. Global surveillance, prevention and control of chronic respiratory diseases: a comprehensive approach [Internet]. 2007 [cited 2016 Apr 1];Available from: http://www.who.int/whr/2007/en/ 2.  Brasier AR, editor. Heterogeneity in Asthma [Internet]. Boston, MA: Springer US; 2014 [cited 2016 Mar 22]. Available from: http://link.springer.com/10.1007/978-1-4614-8603-9 3.  Woodruff PG, Modrek B, Choy DF, Jia G, Abbas AR, Ellwanger A, et al. T-helper Type 2–driven Inflammation Defines Major Subphenotypes of Asthma. Am J Respir Crit Care Med [Internet] 2009 [cited 2016 Mar 27];180:388–95. Available from: http://www.atsjournals.org/doi/abs/10.1164/rccm.200903-0392OC 4.  Wenzel SE. Asthma phenotypes: the evolution from clinical to molecular approaches. Nat Med [Internet] 2012 [cited 2015 Oct 29];18:716–25. Available from: http://www.nature.com.ezproxy.library.ubc.ca/nm/journal/v18/n5/full/nm.2678.html%3FWT.ec_id%3DNM-201205 5.  Global Initiative for Asthma. Global Strategy for Asthma Management and Prevention [Internet]. 2015. Available from: www.ginasthma.org 6.  Wenzel SE. Asthma: defining of the persistent adult phenotypes. The Lancet [Internet] 2006 [cited 2016 Apr 3];368:804–13. Available from: http://www.sciencedirect.com/science/article/pii/S0140673606692908 7.  Levy ML, Quanjer PH, Rachel B, Cooper BG, Holmes S, Small IR. Diagnostic Spirometry in Primary Care: Proposed standards for general practice compliant with American Thoracic Society and European Respiratory Society recommendations. A General Practice Airways Group (GPIAG)1 document, in association with the Association for Respiratory Technology &amp; Physiology (ARTP)2 and Education for Health3 1www.gpiag.org 2 www.artp.org 3 www.educationforhealth.org.uk. Prim Care Respir J [Internet] 2009 [cited 2016 Apr 7];18:130–47. Available from: http://www.nature.com/articles/pcrj200954 8.  Gauvreau GM, Jordana M, Watson RM, Cockcroft DW, O’BYRNE P. Effect of regular inhaled albuterol on allergen-induced late responses and sputum eosinophils in asthmatic subjects. Am J Respir Crit Care Med [Internet] 1997 [cited 2016 Apr 10];156:1738–45. Available from: http://www.atsjournals.org/doi/abs/10.1164/ajrccm.156.6.96-08042 9.  Borish L, Culp JA. Asthma: a syndrome composed of heterogeneous diseases. Ann Allergy Asthma Immunol [Internet] 2008 [cited 2016 Mar 22];101:1–9. Available from: http://www.sciencedirect.com/science/article/pii/S1081120610608265 10.  Craig TJ. Aeroallergen sensitization in asthma: Prevalence and correlation with severity. Allergy Asthma Proc 2010;31:96–102.   156 11.  Malo J-L, Chan-Yeung M. Asthma in the workplace: A Canadian contribution and perspective. Can Respir J 2007;14:407–13.  12.  McFadden Jr ER, Ingram Jr RH. Exercise-induced airway obstruction. Annu Rev Physiol [Internet] 1983 [cited 2016 Apr 12];45:453–63. Available from: http://www.annualreviews.org/doi/pdf/10.1146/annurev.ph.45.030183.002321 13.  Gergen PJ, Turkeltaub PC. The association of allergen skin test reactivity and respiratory disease among whites in the US population: data from the Second National Health and Nutrition Examination Survey, 1976 to 1980. Arch Intern Med [Internet] 1991 [cited 2016 Mar 28];151:487–92. Available from: http://archinte.jamanetwork.com/article.aspx?articleid=614742 14.  Nelson HS, Szefler SJ, Jacobs J, Huss K, Shapiro G, Sternberg AL, et al. The relationships among environmental allergen sensitization, allergen exposure, pulmonary function, and bronchial hyperresponsiveness in the Childhood Asthma Management Program. J Allergy Clin Immunol [Internet] 1999 [cited 2016 Mar 28];104:775–85. Available from: http://www.sciencedirect.com/science/article/pii/S0091674999702873 15.  Duff AL, Pomeranz ES, Gelber LE, Price GW, Farris H, Hayden FG, et al. Risk factors for acute wheezing in infants and children: viruses, passive smoke, and IgE antibodies to inhalant allergens. Pediatrics [Internet] 1993 [cited 2016 Mar 28];92:535–40. Available from: http://pediatrics.aappublications.org/content/92/4/535.short 16.  Contoli M, Ito K, Padovani A, Poletti D, Marku B, Edwards MR, et al. Th2 cytokines impair innate immune responses to rhinovirus in respiratory epithelial cells. Allergy [Internet] 2015 [cited 2016 Mar 28];70:910–20. Available from: http://doi.wiley.com/10.1111/all.12627 17.  Herxheimer H. The Late Bronchial Reaction in Induced Asthma. Int Arch Allergey Immunol 1952;3:323–8.  18.  Booij-Noord H, Orie N, de Vries K. Immediate and late bronchial obstructive reactions to inhalation of house dust and protective effects of disodium cromoglycate and prednisolone. J Allergy Clin Immunol 1971;48:344–54.  19.  Pepys J. Immunopathology of allergic lung disease. Clin Exp Allergy [Internet] 1973 [cited 2016 Apr 3];3:1–22. Available from: http://onlinelibrary.wiley.com/doi/10.1111/j.1365-2222.1973.tb01304.x/full 20.  O’Byrne P, Dolovich J, Hargreave F. Late Asthmatic Responses. Am Rev Respir Dis 1987;136:740–51.  21.  Diamant Z, Gauvreau GM, Cockcroft DW, Boulet L-P, Sterk PJ, de Jongh FHC, et al. Inhaled allergen bronchoprovocation tests. J Allergy Clin Immunol [Internet] 2013 [cited 2015 Jul 10];132:1045–55.e6. Available from: http://linkinghub.elsevier.com/retrieve/pii/S0091674913013377  157 22.  Gauvreau GM, El-Gammal AI, O’Byrne PM. Allergen-induced airway responses. Eur Respir J [Internet] 2015 [cited 2015 Oct 16];46:819–31. Available from: http://erj.ersjournals.com/content/46/3/819 23.  Galli SJ, Tsai M. IgE and mast cells in allergic disease. Nat Med [Internet] 2012 [cited 2016 Apr 4];18:693–704. Available from: http://www.nature.com/doifinder/10.1038/nm.2755 24.  Palaniyandi S, Tomei E, Li Z, Conrad DH, Zhu X. CD23-Dependent Transcytosis of IgE and Immune Complex across the Polarized Human Respiratory Epithelial Cells. J Immunol [Internet] 2011 [cited 2016 Apr 4];186:3484–96. Available from: http://www.jimmunol.org/cgi/doi/10.4049/jimmunol.1002146 25.  Diamant Z, Boot JD, Mantzouranis E, Flohr R, Sterk PJ, Gerth van Wijk R. Biomarkers in asthma and allergic rhinitis. Pulm Pharmacol Ther [Internet] 2010 [cited 2016 Apr 10];23:468–81. Available from: http://linkinghub.elsevier.com/retrieve/pii/S1094553910000799 26.  Hamilton A, Faiferman I, Stober P, Watson RM, O’Byrne PM. Pranlukast, a cysteinyl leukotriene receptor antagonist, attenuates allergen-induced early-and late-phase bronchoconstriction and airway hyperresponsiveness in asthmatic subjects. J Allergy Clin Immunol [Internet] 1998 [cited 2016 Apr 10];102:177–83. Available from: http://www.sciencedirect.com/science/article/pii/S0091674998700831 27.  Kouro T, Takatsu K. IL-5- and eosinophil-mediated inflammation: from discovery to therapy. Int Immunol [Internet] 2009 [cited 2016 Apr 10];21:1303–9. Available from: http://www.intimm.oxfordjournals.org/cgi/doi/10.1093/intimm/dxp102 28.  Samitas K, Delimpoura V, Zervas E, Gaga M. Anti-IgE treatment, airway inflammation and remodelling in severe allergic asthma: current knowledge and future perspectives. Eur Respir Rev [Internet] 2015 [cited 2016 Apr 7];24:594–601. Available from: http://err.ersjournals.com/lookup/doi/10.1183/16000617.00001715 29.  GAUVREAU GM, LEE JM, WATSON RM, IRANI A-MA, SCHWARTZ LB, O’BYRNE PM. Increased numbers of both airway basophils and mast cells in sputum after allergen inhalation challenge of atopic asthmatics. Am J Respir Crit Care Med [Internet] 2000 [cited 2016 Mar 22];161:1473–8. Available from: http://www.atsjournals.org/doi/abs/10.1164/ajrccm.161.5.9908090 30.  Matsumoto K, Gauvreau GM, Rerecich T, Watson RM, Wood LJ, O’Byrne PM. IL-10 production in circulating T cells differs between allergen-induced isolated early and dual asthmatic responders. J Allergy Clin Immunol [Internet] 2002 [cited 2016 Apr 5];109:281–6. Available from: http://linkinghub.elsevier.com/retrieve/pii/S0091674902554932 31.  Hammad H, Chieppa M, Perros F, Willart MA, Germain RN, Lambrecht BN. House dust mite allergen induces asthma via Toll-like receptor 4 triggering of airway structural cells. Nat Med [Internet] 2009 [cited 2016 Apr 11];15:410–6. Available from: http://www.nature.com/doifinder/10.1038/nm.1946  158 32.  Claudio E, Tassi I, Wang H, Tang W, Ha H -l., Siebenlist U. Cutting Edge: IL-25 Targets Dendritic Cells To Attract IL-9-Producing T Cells in Acute Allergic Lung Inflammation. J Immunol [Internet] 2015 [cited 2016 Apr 11];195:3525–9. Available from: http://www.jimmunol.org/cgi/doi/10.4049/jimmunol.1500436 33.  Lambrecht BN, Hammad H. The airway epithelium in asthma. Nat Med [Internet] 2012 [cited 2016 Apr 11];18:684–92. Available from: http://www.nature.com/doifinder/10.1038/nm.2737 34.  Dahl R. Systemic side effects of inhaled corticosteroids in patients with asthma. Respir Med [Internet] 2006 [cited 2016 Apr 11];100:1307–17. Available from: http://linkinghub.elsevier.com/retrieve/pii/S095461110500510X 35.  Gauvreau GM, Boulet L-P, Leigh R, Cockcroft DW, Killian KJ, Davis BE, et al. A Nonsteroidal Glucocorticoid Receptor Agonist Inhibits Allergen-induced Late Asthmatic Responses. Am J Respir Crit Care Med [Internet] 2014 [cited 2015 Oct 29];191:161–7. Available from: http://www.atsjournals.org.ezproxy.library.ubc.ca/doi/abs/10.1164/rccm.201404-0623OC 36.  Yick CY, Zwinderman AH, Kunst PW, Grunberg K, Mauad T, Dijkhuis A, et al. Transcriptome sequencing (RNA-Seq) of human endobronchial biopsies: asthma versus controls. Eur Respir J [Internet] 2013 [cited 2016 Mar 29];42:662–70. Available from: http://erj.ersjournals.com/cgi/doi/10.1183/09031936.00115412 37.  Baines KJ, Simpson JL, Wood LG, Scott RJ, Gibson PG. Transcriptional phenotypes of asthma defined by gene expression profiling of induced sputum samples. J Allergy Clin Immunol [Internet] 2011 [cited 2016 Mar 29];127:153–60.e9. Available from: http://linkinghub.elsevier.com/retrieve/pii/S0091674910016386 38.  Mattila P, Renkonen J, Toppila-Salmi S, Parviainen V, Joenväärä S, Alff-Tuomala S, et al. Time-series nasal epithelial transcriptomics during natural pollen exposure in healthy subjects and allergic patients. Allergy [Internet] 2010 [cited 2016 Mar 29];65:175–83. Available from: http://doi.wiley.com/10.1111/j.1398-9995.2009.02181.x 39.  Carlsten C, Blomberg A, Pui M, Sandstrom T, Wong SW, Alexis N, et al. Diesel exhaust augments allergen-induced lower airway inflammation in allergic individuals: a controlled human exposure study. Thorax [Internet] 2015 [cited 2016 Mar 29];thoraxjnl – 2015–207399. Available from: http://thorax.bmj.com/lookup/doi/10.1136/thoraxjnl-2015-207399 40.  Himes BE, Jiang X, Wagner P, Hu R, Wang Q, Klanderman B, et al. RNA-Seq Transcriptome Profiling Identifies CRISPLD2 as a Glucocorticoid Responsive Gene that Modulates Cytokine Function in Airway Smooth Muscle Cells. PLoS ONE [Internet] 2014 [cited 2016 Mar 29];9:e99625. Available from: http://dx.plos.org/10.1371/journal.pone.0099625 41.  Kicic A, Hallstrand TS, Sutanto EN, Stevens PT, Kobor MS, Taplin C, et al. Decreased Fibronectin Production Significantly Contributes to Dysregulated Repair of Asthmatic  159 Epithelium. Am J Respir Crit Care Med [Internet] 2010 [cited 2016 Apr 11];181:889–98. Available from: http://www.atsjournals.org/doi/abs/10.1164/rccm.200907-1071OC 42.  Woodruff PG, Boushey HA, Dolganov GM, Barker CS, Yang YH, Donnelly S, et al. Genome-wide profiling identifies epithelial cell genes associated with asthma and with treatment response to corticosteroids. Proc Natl Acad Sci [Internet] 2007 [cited 2016 Mar 29];104:15858–63. Available from: http://www.pnas.org/content/104/40/15858.short 43.  Subrata LS, Bizzintino J, Mamessier E, Bosco A, McKenna KL, Wikstrom ME, et al. Interactions between Innate Antiviral and Atopic Immunoinflammatory Pathways Precipitate and Sustain Asthma Exacerbations in Children. J Immunol [Internet] 2009 [cited 2015 Nov 19];183:2793–800. Available from: http://www.jimmunol.org/cgi/doi/10.4049/jimmunol.0900695 44.  Bjornsdottir US, Holgate ST, Reddy PS, Hill AA, McKee CM, Csimma CI, et al. Pathways Activated during Human Asthma Exacerbation as Revealed by Gene Expression Patterns in Blood. PLoS ONE [Internet] 2011 [cited 2015 Nov 19];6:e21902. Available from: http://dx.plos.org/10.1371/journal.pone.0021902 45.  Terracciano R, Pelaia G, Preianò M, Savino R. Asthma and COPD proteomics: Current approaches and future directions. PROTEOMICS - Clin Appl [Internet] 2015 [cited 2016 Apr 12];9:203–20. Available from: http://doi.wiley.com/10.1002/prca.201400099 46.  Zhao J, Endoh I, Hsu K, Tedla N, Endoh Y, Geczy CL. S100A8 Modulates Mast Cell Function and Suppresses Eosinophil Migration in Acute Asthma. Antioxid Redox Signal 2011;14:1589–600.  47.  Gharib SA, Nguyen EV, Lai Y, Plampin JD, Goodlett DR, Hallstrand TS. Induced sputum proteome in healthy subjects and asthmatic patients. J Allergy Clin Immunol [Internet] 2011 [cited 2016 Apr 12];128:1176–84.e6. Available from: http://linkinghub.elsevier.com/retrieve/pii/S0091674911012474 48.  Wu J, Kobayashi M, Sousa EA, Liu W, Cai J, Goldman SJ, et al. Differential proteomic analysis of bronchoalveolar lavage fluid in asthmatics following segmental antigen challenge. Mol Cell Proteomics [Internet] 2005 [cited 2016 Mar 23];4:1251–64. Available from: http://www.mcponline.org/content/4/9/1251.short 49.  Nishioka T, Uchida K, Meno K, Ishii T, Aoki T, Imada Y, et al. Alpha-1-antitrypsin and complement component C7 are involved in asthma exacerbation. PROTEOMICS – Clin Appl [Internet] 2008 [cited 2016 Apr 12];2:46–54. Available from: http://doi.wiley.com/10.1002/prca.200780065 50.  Rhim T, Choi Y-S, Nam B-Y, Uh ST, Park JS, Kim Y-H, et al. Plasma protein profiles in early asthmatic responses to inhalation allergen challenge. Allergy [Internet] 2009 [cited 2016 Mar 23];64:47–54. Available from: http://doi.wiley.com/10.1111/j.1398-9995.2008.01817.x  160 51.  Adamko DJ, Sykes BD, Rowe BH. The Metabolomics of Asthma. Chest [Internet] 2012 [cited 2016 Apr 12];141:1295–302. Available from: http://linkinghub.elsevier.com/retrieve/pii/S0012369212602888 52.  Reisdorph N, Wechsler ME. Utilizing metabolomics to distinguish asthma phenotypes: strategies and clinical implications. Allergy [Internet] 2013 [cited 2016 Apr 12];68:959–62. Available from: http://doi.wiley.com/10.1111/all.12238 53.  Saude EJ, Skappak CD, Regush S, Cook K, Ben-Zvi A, Becker A, et al. Metabolomic profiling of asthma: Diagnostic utility of urine nuclear magnetic resonance spectroscopy. J Allergy Clin Immunol [Internet] 2011 [cited 2016 Apr 12];127:757–64.e6. Available from: http://linkinghub.elsevier.com/retrieve/pii/S0091674910030186 54.  Jung J, Kim S-H, Lee H-S, Choi GS, Jung Y-S, Ryu DH, et al. Serum metabolomics reveals pathways and biomarkers associated with asthma pathogenesis. Clin Exp Allergy [Internet] 2013 [cited 2016 Apr 12];43:425–33. Available from: http://doi.wiley.com/10.1111/cea.12089 55.  Ho WE, Xu Y-J, Xu F, Cheng C, Peh HY, Tannenbaum SR, et al. Metabolomics Reveals Altered Metabolic Pathways in Experimental Asthma. Am J Respir Cell Mol Biol [Internet] 2013 [cited 2016 Apr 12];48:204–11. Available from: http://www.atsjournals.org/doi/abs/10.1165/rcmb.2012-0246OC 56.  Ibrahim B, Marsden P, Smith JA, Custovic A, Nilsson M, Fowler SJ. Breath metabolomic profiling by nuclear magnetic resonance spectroscopy in asthma. Allergy [Internet] 2013 [cited 2016 Apr 12];68:1050–6. Available from: http://doi.wiley.com/10.1111/all.12211 57.  Szefler SJ, Wenzel S, Brown R, Erzurum SC, Fahy JV, Hamilton RG, et al. Asthma outcomes: Biomarkers. J Allergy Clin Immunol [Internet] 2012 [cited 2015 Nov 30];129:S9–23. Available from: http://linkinghub.elsevier.com/retrieve/pii/S0091674911029484 58.  Fitzpatrick AM. Biomarkers of asthma and allergic airway diseases. Ann Allergy Asthma Immunol [Internet] 2015 [cited 2016 Mar 29];115:335–40. Available from: http://linkinghub.elsevier.com/retrieve/pii/S108112061500589X 59.  See KC, Christiani DC. Normal values and thresholds for the clinical interpretation of exhaled nitric oxide levels in the US general population: results from the National Health and Nutrition Examination Survey 2007-2010. CHEST J [Internet] 2013 [cited 2016 Apr 7];143:107–16. Available from: http://journal.publications.chestnet.org/article.aspx?articleid=1216074 60.  Jia G, Erickson RW, Choy DF, Mosesova S, Wu LC, Solberg OD, et al. Periostin is a systemic biomarker of eosinophilic airway inflammation in asthmatic patients. J Allergy Clin Immunol [Internet] 2012 [cited 2016 Mar 29];130:647–54.e10. Available from: http://linkinghub.elsevier.com/retrieve/pii/S0091674912010299 61.  Corren J, Lemanske Jr RF, Hanania NA, Korenblat PE, Parsey MV, Arron JR, et al. Lebrikizumab treatment in adults with asthma. N Engl J Med [Internet] 2011 [cited 2016  161 Mar 29];365:1088–98. Available from: http://www.nejm.org/doi/full/10.1056/Nejmoa1106469 62.  Tang H, Fang Z, Sun Y, Li B, Shi Z, Chen J, et al. YKL-40 in asthmatic patients, and its correlations with exacerbation, eosinophils and immunoglobulin E. Eur Respir J [Internet] 2010 [cited 2016 Apr 12];35:757–60. Available from: http://erj.ersjournals.com/cgi/doi/10.1183/09031936.00034409 63.  Rabinovitch N, Zhang L, Gelfand E. Urine leukotriene E4 levels are associated with decreased pulmonary function in children with persistent airway obstruction. J Allergy Clin Immunol [Internet] 2006 [cited 2016 Apr 12];118:635–40. Available from: http://linkinghub.elsevier.com/retrieve/pii/S0091674906011456 64.  Green S, Malice M-P, Tanaka W, Tozzi C, Reiss T. Increase in urinary leukotriene LTE4 levels in acute asthma: correlation with airflow limitation. Thorax 2004;59:100–4.  65.  Dharajiya N, Vaidya S, Sinha M, Luxon B, Boldogh I, Sur S. Allergen Challenge Induces Ifng Dependent GTPases in the Lungs as Part of a Th1 Transcriptome Response in a Murine Model of Allergic Asthma. PLoS ONE [Internet] 2009 [cited 2016 Mar 22];4:e8172. Available from: http://dx.plos.org/10.1371/journal.pone.0008172 66.  Zou J, Young S, Zhu F, Gheyas F, Skeans S, Wan Y, et al. Microarray profile of differentially expressed genes in a monkey model of allergic asthma. Genome Biol [Internet] 2002 [cited 2016 Mar 22];3:20. Available from: http://www.biomedcentral.com/content/pdf/gb-2002-3-5-research0020.pdf 67.  Bailey MT, Kierstein S, Sharma S, Spaits M, Kinsey SG, Tliba O, et al. Social Stress Enhances Allergen-Induced Airway Inflammation in Mice and Inhibits Corticosteroid Responsiveness of Cytokine Production. J Immunol [Internet] 2009 [cited 2016 Mar 22];182:7888–96. Available from: http://www.jimmunol.org/cgi/doi/10.4049/jimmunol.0800891 68.  Frew AJ, Moqbel R, Azzawi M, Hartnell A, Barkans J, Jeffery PK, et al. T lymphocytes and eosinophils in allergen-induced late-phase asthmatic reactions in the guinea pig. Am Rev Respir Dis [Internet] 1990 [cited 2015 Oct 29];141:407–13. Available from: http://www.atsjournals.org.ezproxy.library.ubc.ca/doi/abs/10.1164/ajrccm/141.2.407 69.  Corrigan CJ, Wang W, Meng Q, Fang C, Eid G, Caballero MR, et al. Allergen-induced expression of IL-25 and IL-25 receptor in atopic asthmatic airways and late-phase cutaneous responses. J Allergy Clin Immunol [Internet] 2011 [cited 2015 Oct 29];128:116–24. Available from: http://www.sciencedirect.com/science/article/pii/S0091674911005744 70.  El-Gammal A, Oliveria J-P, Howie K, Watson R, Mitchell P, Chen R, et al. Allergen-induced Changes in Bone Marrow and Airway Dendritic Cells in Asthmatic Subjects. Am J Respir Crit Care Med [Internet] 2016 [cited 2016 Mar 22];Available from: http://www.atsjournals.org/doi/10.1164/rccm.201508-1623OC  162 71.  Yoshida M, Watson R, Rerecich T, Obyrne P. Different profiles of T-cell IFN-γ and IL-12 in allergen-induced early and dual responders with asthma. J Allergy Clin Immunol [Internet] 2005 [cited 2016 Mar 22];115:1004–9. Available from: http://linkinghub.elsevier.com/retrieve/pii/S0091674905003386 72.  Kinoshita T, Baatjes A, Smith SG, Dua B, Watson R, Kawayama T, et al. Natural regulatory T cells in isolated early responders compared with dual responders with allergic asthma. J Allergy Clin Immunol [Internet] 2014 [cited 2016 Mar 22];133:696–703. Available from: http://linkinghub.elsevier.com/retrieve/pii/S0091674913013390 73.  Naji N, Smith S, Gauvreau G, O’Byrne P. T Helper 17 Cells and Related Cytokines after Allergen Inhalation Challenge in Allergic Asthmatics. Int Arch Allergey Immunol 2014;165.  74.  Bartel DP. MicroRNAs: Target Recognition and Regulatory Functions. Cell [Internet] 2009 [cited 2016 Mar 22];136:215–33. Available from: http://linkinghub.elsevier.com/retrieve/pii/S0092867409000087 75.  Garbacki N, Di Valentin E, Huynh-Thu VA, Geurts P, Irrthum A, Crahay C, et al. MicroRNAs Profiling in Murine Models of Acute and Chronic Asthma: A Relationship with mRNAs Targets. PLoS ONE [Internet] 2011 [cited 2016 Mar 22];6:e16509. Available from: http://dx.plos.org/10.1371/journal.pone.0016509 76.  Lu TX, Munitz A, Rothenberg ME. MicroRNA-21 Is Up-Regulated in Allergic Airway Inflammation and Regulates IL-12p35 Expression. J Immunol [Internet] 2009 [cited 2016 Mar 22];182:4994–5002. Available from: http://www.jimmunol.org/cgi/doi/10.4049/jimmunol.0803560 77.  Simpson LJ, Patel S, Bhakta NR, Choy DF, Brightbill HD, Ren X, et al. A microRNA upregulated in asthma airway T cells promotes TH2 cytokine production. Nat Immunol [Internet] 2014 [cited 2016 Mar 22];15:1162–70. Available from: http://www.nature.com/doifinder/10.1038/ni.3026 78.  Greene CM, Gaughan KP. microRNAs in asthma: potential therapeutic targets. Curr Opin Pulm Med [Internet] 2013 [cited 2016 Mar 22];19:66–72. Available from: http://content.wkhealth.com/linkback/openurl?sid=WKPTLP:landingpage&an=00063198-201301000-00013 79.  O’Neil SE, Lundbäck B, Lötvall J. Proteomics in asthma and COPD phenotypes and endotypes for biomarker discovery and improved understanding of disease entities. J Proteomics [Internet] 2011 [cited 2016 Mar 22];75:192–201. Available from: http://linkinghub.elsevier.com/retrieve/pii/S1874391911004945 80.  Haenen S, Clynen E, Nemery B, Hoet PHM, Vanoirbeek JAJ. Biomarker discovery in asthma and COPD: Application of proteomics techniques in human and mice. EuPA Open Proteomics [Internet] 2014 [cited 2016 Mar 22];4:101–12. Available from: http://linkinghub.elsevier.com/retrieve/pii/S2212968514000336  163 81.  Cockcroft DW, Murdock KY, Kirby J, Hargreave F. Prediction of Airway Responsiveness to Allergen from Skin Sensitivity to Allergen and Airway Responsiveness to Histamine. Am Rev Respir Dis 1987;135:264–7.  82.  Geiss GK, Bumgarner RE, Birditt B, Dahl T, Dowidar N, Dunaway DL, et al. Direct multiplexed measurement of gene expression with color-coded probe pairs. Nat Biotechnol [Internet] 2008 [cited 2015 Dec 16];26:317–25. Available from: http://www.nature.com/doifinder/10.1038/nbt1385 83.  Kuzyk MA, Ohlund LB, Elliott MH, Smith D, Qian H, Delaney A, et al. A comparison of MS/MS-based, stable-isotope-labeled, quantitation performance on ESI-quadrupole TOF and MALDI-TOF/TOF mass spectrometers. PROTEOMICS [Internet] 2009 [cited 2016 Mar 15];9:3328–40. Available from: http://doi.wiley.com/10.1002/pmic.200800412 84.  Kuzyk MA, Smith D, Yang J, Cross TJ, Jackson AM, Hardie DB, et al. Multiple reaction monitoring-based, multiplexed, absolute quantitation of 45 proteins in human plasma. Mol Cell Proteomics [Internet] 2009 [cited 2016 Mar 15];8:1860–77. Available from: http://www.mcponline.org/content/8/8/1860.short 85.  Hochreiter S, Clevert D-A, Obermayer K. A new summarization method for affymetrix probe level data. Bioinformatics [Internet] 2006 [cited 2016 Mar 11];22:943–9. Available from: http://bioinformatics.oxfordjournals.org/cgi/doi/10.1093/bioinformatics/btl033 86.  Talloen W, Clevert D-A, Hochreiter S, Amaratunga D, Bijnens L, Kass S, et al. I/NI-calls for the exclusion of non-informative genes: a highly effective filtering tool for microarray data. Bioinformatics [Internet] 2007 [cited 2016 Mar 11];23:2897–902. Available from: http://bioinformatics.oxfordjournals.org/cgi/doi/10.1093/bioinformatics/btm478 87.  Shilov IV, Seymour SL, Patel AA, Loboda A, Tang WH, Keating SP, et al. The Paragon Algorithm, a next generation search engine that uses sequence temperature values and feature probabilities to identify peptides from tandem mass spectra. Mol Cell Proteomics [Internet] 2007 [cited 2016 Mar 15];6:1638–55. Available from: https://www.mcponline.org/content/6/9/1638.full 88.  Kersey PJ, Duarte J, Williams A, Karavidopoulou Y, Birney E, Apweiler R. The International Protein Index: An integrated database for proteomics experiments. PROTEOMICS [Internet] 2004 [cited 2016 Mar 15];4:1985–8. Available from: http://doi.wiley.com/10.1002/pmic.200300721 89.  Ritchie ME, Phipson B, Wu D, Hu Y, Law CW, Shi W, et al. limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res [Internet] 2015 [cited 2016 Mar 15];43:e47–e47. Available from: http://nar.oxfordjournals.org/lookup/doi/10.1093/nar/gkv007 90.  Smyth G. Linear Models and Empirical Bayes Methods for Assessing Differential Expression in Microarray Experiments. Stat Appl Genet Mol Biol 2004;3:26.   164 91.  Benjamini Y, Hochberg Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J R Stat Soc Ser B Methodol [Internet] 1995 [cited 2016 Mar 11];289–300. Available from: http://www.jstor.org/stable/2346101 92.  Shen-Orr SS, Tibshirani R, Khatri P, Bodian DL, Staedtler F, Perry NM, et al. Cell type–specific gene expression differences in complex tissues. Nat Methods [Internet] 2010 [cited 2016 Mar 15];7:287–9. Available from: http://www.nature.com/doifinder/10.1038/nmeth.1439 93.  Pearson K. On lines and planes of closest fit to systems of points in space. Philos Mag 1901;6:559–72.  94.  Hotelling H. Analysis of a complex of statistical variables into principal components. J Educ Psychol [Internet] 1933 [cited 2016 Mar 16];24:417. Available from: http://psycnet.apa.org/journals/edu/24/6/417/ 95.  R Core Team. R: A Language and Environment for Statistical Computing [Internet]. Vienna, Austria: R Foundation for Statistical Computing; 2015. Available from: https://www.R-project.org/ 96.  Lê Cao K-A, Boitard S, Besse P. Sparse PLS discriminant analysis: biologically relevant feature selection and graphical displays for multiclass problems. BMC Bioinformatics [Internet] 2011 [cited 2015 Jul 15];12:253. Available from: http://www.biomedcentral.com/1471-2105/12/253/ 97.  Izzotti A, Calin GA, Arrigo P, Steele VE, Croce CM, De Flora S. Downregulation of microRNA expression in the lungs of rats exposed to cigarette smoke. FASEB J [Internet] 2009 [cited 2016 Mar 23];23:806–12. Available from: http://www.fasebj.org/cgi/doi/10.1096/fj.08-121384 98.  Song B, Wang Y, Kudo K, Gavin EJ, Xi Y, Ju J. miR-192 Regulates Dihydrofolate Reductase and Cellular Proliferation through the p53-microRNA Circuit. Clin Cancer Res [Internet] 2008 [cited 2016 Mar 23];14:8080–6. Available from: http://clincancerres.aacrjournals.org/cgi/doi/10.1158/1078-0432.CCR-08-1422 99.  Wang G, Tam L, Li E, Kwan B, Chow K, Luk C, et al. Serum and urinary free microRNA level in patients with systemic lupus erythematosus. Lupus [Internet] 2011 [cited 2016 Mar 23];20:493–500. Available from: http://lup.sagepub.com/cgi/doi/10.1177/0961203310389841 100.  Krupa A, Jenkins R, Luo DD, Lewis A, Phillips A, Fraser D. Loss of MicroRNA-192 Promotes Fibrogenesis in Diabetic Nephropathy. J Am Soc Nephrol [Internet] 2010 [cited 2016 Mar 23];21:438–47. Available from: http://www.jasn.org/cgi/doi/10.1681/ASN.2009050530 101.  Torrego A, Hew M, Oates T, Sukkar M, Fan Chung K. Expression and activation of TGF- isoforms in acute allergen-induced remodelling in asthma. Thorax [Internet] 2007  165 [cited 2016 Mar 23];62:307–13. Available from: http://thorax.bmj.com/cgi/doi/10.1136/thx.2006.063487 102.  Rossi RL, Rossetti G, Wenandy L, Curti S, Ripamonti A, Bonnal RJP, et al. Distinct microRNA signatures in human lymphocyte subsets and enforcement of the naive state in CD4+ T cells by the microRNA miR-125b. Nat Immunol [Internet] 2011 [cited 2016 Mar 23];12:796–803. Available from: http://www.nature.com/doifinder/10.1038/ni.2057 103.  Gerblich A, Salik H, Schuyler M. Dynamic T-Cell Changes in Peripheral Blood and Bronchoalveolar Lavage after Antigen Bronchoprovocation in Asthmatics. Am J Respir Crit Care Med 1991;143:533–7.  104.  Verrills NM, Irwin JA, Yan He X, Wood LG, Powell H, Simpson JL, et al. Identification of Novel Diagnostic Biomarkers for Asthma and Chronic Obstructive Pulmonary Disease. Am J Respir Crit Care Med [Internet] 2011 [cited 2016 Mar 23];183:1633–43. Available from: http://www.atsjournals.org/doi/abs/10.1164/rccm.201010-1623OC 105.  Yamada K, Hasegawa T, Hasegawa E, Kennedy D, Hirano H, Hayashi M, et al. Fibronectin and interactions at the cell surface. Prog Clin Biol Res 1984;151:1–15.  106.  Eriksen HO, Clemmensen I, Hansen MS, Ibsen KK. Plasma fibronectin concentration in normal subjects. Scand J Clin Lab Invest [Internet] 1982 [cited 2016 Mar 23];42:291–5. Available from: http://www.tandfonline.com/doi/full/10.1080/00365518209168088 107.  Hocking D. Fibronectin matrix deposition and cell contractility: implications for airway remodeling in asthma. Chest 2002;122:275S – 278S.  108.  Roche W, Williams J, Beasley R, Holgate S. Subepithelial fibrosis in the bronchi of asthmatics. The Lancet 1989;333:520–4.  109.  Slats AM, Janssen K, van Schadewijk A, van der Plas DT, Schot R, van den Aardweg JG, et al. Expression of smooth muscle and extracellular matrix proteins in relation to airway function in asthma. J Allergy Clin Immunol [Internet] 2008 [cited 2016 Mar 23];121:1196–202. Available from: http://linkinghub.elsevier.com/retrieve/pii/S0091674908003667 110.  Yick CY, Ferreira DS, Annoni R, Thüsen JH, Kunst PW, Bel EH, et al. Extracellular matrix in airway smooth muscle is associated with dynamics of airway function in asthma. Allergy [Internet] 2012 [cited 2016 Mar 23];67:552–9. Available from: http://doi.wiley.com/10.1111/j.1398-9995.2011.02773.x 111.  Domanski D, Percy AJ, Yang J, Chambers AG, Hill JS, Freue GVC, et al. MRM-based multiplexed quantitation of 67 putative cardiovascular disease biomarkers in human plasma. PROTEOMICS [Internet] 2012 [cited 2016 Mar 23];12:1222–43. Available from: http://doi.wiley.com/10.1002/pmic.201100568 112.  Meng C, Kuster B, Culhane AC, Gholami AM. A multivariate approach to the integration of multi-omics datasets. BMC Bioinformatics [Internet] 2014 [cited 2016 Jan 19];15:162.  166 Available from: http://www.biomedcentral.com/1471-2105/15/162?utm_source=dlvr.it&utm_medium=tumblr 113.  Liu Y, Devescovi V, Chen S, Nardini C. Multilevel omic data integration in cancer cell lines: advanced annotation and emergent properties. BMC Syst Biol [Internet] 2013 [cited 2016 Jan 19];7:14. Available from: http://www.biomedcentral.com/1752-0509/7/14 114.  Kirk P, Griffin JE, Savage RS, Ghahramani Z, Wild DL. Bayesian correlated clustering to integrate multiple datasets. Bioinformatics [Internet] 2012 [cited 2016 Jan 19];28:3290–7. Available from: http://bioinformatics.oxfordjournals.org/cgi/doi/10.1093/bioinformatics/bts595 115.  Zhu J, Sova P, Xu Q, Dombek KM, Xu EY, Vu H, et al. Stitching together Multiple Data Dimensions Reveals Interacting Metabolomic and Transcriptomic Networks That Modulate Cell Regulation. PLoS Biol [Internet] 2012 [cited 2016 Jan 19];10:e1001301. Available from: http://dx.plos.org/10.1371/journal.pbio.1001301 116.  Wang B, Mezlini AM, Demir F, Fiume M, Tu Z, Brudno M, et al. Similarity network fusion for aggregating data types on a genomic scale. Nat Methods [Internet] 2014 [cited 2016 Jan 19];11:333–7. Available from: http://www.nature.com/doifinder/10.1038/nmeth.2810 117.  Glass K, Huttenhower C, Quackenbush J, Yuan G-C. Passing Messages between Biological Networks to Refine Predicted Interactions. PLoS ONE [Internet] 2013 [cited 2015 Dec 1];8:e64832. Available from: http://dx.plos.org/10.1371/journal.pone.0064832 118.  Sharma S, Poon A, Himes BE, Lasky-Su J, Sordillo JE, Belanger K, et al. Association of variants in innate immune genes with asthma and eczema: Innate immune genes and atopic disease. Pediatr Allergy Immunol [Internet] 2012 [cited 2016 Mar 30];23:315–23. Available from: http://doi.wiley.com/10.1111/j.1399-3038.2011.01243.x 119.  Sanders DL, Aronsky D. Prospective Evaluation of a Bayesian Network for Detecting Asthma Exacerbations in a Pediatric Emergency Department. In: AMIA Symposium. 2006. page 1. 120.  Himes BE, Wu AC, Duan QL, Klanderman B, Litonjua AA, Tantisira K, et al. Predicting response to short-acting bronchodilator medication using Bayesian networks. Pharmacogenomics [Internet] 2009 [cited 2016 Mar 30];10:1393–412. Available from: http://www.futuremedicine.com/doi/abs/10.2217/pgs.09.93 121.  Xu M, Tantisira KG, Wu A, Litonjua AA, Chu J, Himes BE, et al. Genome Wide Association Study to predict severe asthma exacerbations in children using random forests classifiers. BMC Med Genet [Internet] 2011 [cited 2016 Mar 30];12:1. Available from: http://bmcmedgenet.biomedcentral.com/articles/10.1186/1471-2350-12-90 122.  Bunyavanich S, Schadt EE. Systems biology of asthma and allergic diseases: A multiscale approach. J Allergy Clin Immunol [Internet] 2015 [cited 2015 Nov 29];135:31–42. Available from: http://linkinghub.elsevier.com/retrieve/pii/S0091674914014869  167 123.  Sehouli J, Loddenkemper C, Cornu T, Schwachula T, Hoffmüller U, Grützkau A, et al. Epigenetic quantification of tumor-infiltrating T-lymphocytes. Epigenetics [Internet] 2011 [cited 2016 Mar 17];6:236–46. Available from: http://www.tandfonline.com/doi/abs/10.4161/epi.6.2.13755 124.  Wieczorek G, Asemissen A, Model F, Turbachova I, Floess S, Liebenberg V, et al. Quantitative DNA Methylation Analysis of FOXP3 as a New Method for Counting Regulatory T Cells in Peripheral Blood and Solid Tissue. Cancer Res [Internet] 2009 [cited 2016 Mar 17];69:599–608. Available from: http://cancerres.aacrjournals.org/cgi/doi/10.1158/0008-5472.CAN-08-2361 125.  Olek A, Oswald J, Walter J. A modified and improved method for bisulphite based cytosine methylation analysis. Nucleic Acids Res [Internet] 1996 [cited 2016 Mar 17];24:5064–6. Available from: https://nar.oxfordjournals.org/content/24/24/5064.full 126.  Folch J, Lees M, Sloane-Stanley GH, others. A simple method for the isolation and purification of total lipids from animal tissues. J Biol Chem [Internet] 1957 [cited 2016 Apr 2];226:497–509. Available from: http://www.aufsi.auburn.edu/recommendedmethods/05B01c03a.pdf 127.  Schlenk H, Gellerman JL. Esterification of Fatty Acids with Diazomethane on a Small Scale. Anal Chem 1960;32:1412–4.  128.  González I, Déjean S, Martin PG, Gonçalves O, Besse P, Baccini A. Highlighting relationships between heterogeneous biological data through graphical displays based on regularized canonical correlation analysis. J Biol Syst [Internet] 2009 [cited 2016 Mar 17];17:173–99. Available from: http://www.worldscientific.com/doi/abs/10.1142/S0218339009002831 129.  Le Cao K-A, Gonzalez I, Dejean S. integrOmics: an R package to unravel relationships between two omics datasets. Bioinformatics [Internet] 2009 [cited 2016 Apr 3];25:2855–6. Available from: http://bioinformatics.oxfordjournals.org/cgi/doi/10.1093/bioinformatics/btp515 130.  González I, Lê Cao K-A, Davis MJ, Déjean S. Visualising associations between paired “omics” data sets. BioData Min [Internet] 2012 [cited 2015 Jul 15];5:1–23. Available from: http://link.springer.com/article/10.1186/1756-0381-5-19 131.  Hamann J, Koning N, Pouwels W, Ulfman LH, van Eijk M, Stacey M, et al. EMR1, the human homolog of F4/80, is an eosinophil-specific receptor. Eur J Immunol [Internet] 2007 [cited 2016 Apr 2];37:2797–802. Available from: http://onlinelibrary.wiley.com/doi/10.1002/eji.200737553/full 132.  Seeds MC, Peachman KK, Bowton DL, Sivertson KL, Chilton FH. Regulation of Arachidonate Remodeling Enzymes Impacts Eosinophil Survival during Allergic Asthma. Am J Respir Cell Mol Biol [Internet] 2009 [cited 2016 Apr 2];41:358–66. Available from: http://www.atsjournals.org/doi/abs/10.1165/rcmb.2008-0192OC  168 133.  Rzehak P, Thijs C, Standl M, Mommers M, Glaser C, Jansen E, et al. Variants of the FADS1 FADS2 Gene Cluster, Blood Levels of Polyunsaturated Fatty Acids and Eczema in Children within the First 2 Years of Life. PLoS ONE [Internet] 2010 [cited 2016 Apr 2];5:e13261. Available from: http://dx.plos.org/10.1371/journal.pone.0013261 134.  Serhan CN, Chiang N, Van Dyke TE. Resolving inflammation: dual anti-inflammatory and pro-resolution lipid mediators. Nat Rev Immunol [Internet] 2008 [cited 2016 Apr 2];8:349–61. Available from: http://www.nature.com/doifinder/10.1038/nri2294 135.  Chiang N, Arita M, Serhan CN. Anti-inflammatory circuitry: Lipoxin, aspirin-triggered lipoxins and their receptor ALX. Prostaglandins Leukot Essent Fatty Acids [Internet] 2005 [cited 2016 Apr 2];73:163–77. Available from: http://linkinghub.elsevier.com/retrieve/pii/S0952327805000815 136.  Levy BD, Kohli P, Gotlinger K, Haworth O, Hong S, Kazani S, et al. Protectin D1 Is Generated in Asthma and Dampens Airway Inflammation and Hyperresponsiveness. J Immunol [Internet] 2007 [cited 2016 Apr 2];178:496–502. Available from: http://www.jimmunol.org/cgi/doi/10.4049/jimmunol.178.1.496 137.  Feltenmark S, Gautam N, Brunnström \AAsa, Griffiths W, Backman L, Edenius C, et al. Eoxins are proinflammatory arachidonic acid metabolites produced via the 15-lipoxygenase-1 pathway in human eosinophils and mast cells. Proc Natl Acad Sci [Internet] 2008 [cited 2016 Apr 2];105:680–5. Available from: https://www.pnas.org/content/105/2/680.full 138.  Levy BD. Resolvins and protectins: Natural pharmacophores for resolution biology. Prostaglandins Leukot Essent Fat Acids PLEFA [Internet] 2010 [cited 2016 Apr 2];82:327–32. Available from: http://linkinghub.elsevier.com/retrieve/pii/S0952327810000475 139.  Perogamvros I, Ray DW, Trainer PJ. Regulation of cortisol bioavailability—effects on hormone measurement and action. Nat Rev Endocrinol [Internet] 2012 [cited 2016 Apr 2];8:717–27. Available from: http://www.nature.com/doifinder/10.1038/nrendo.2012.134 140.  Barnes PJ. Bradykinin and asthma. Thorax [Internet] 1992 [cited 2016 Apr 2];47:979–83. Available from: http://thorax.bmj.com/content/47/11/979.full.pdf 141.  Levy BD, Vachier I, Serhan CN. Resolution of Inflammation in Asthma. Clin Chest Med [Internet] 2012 [cited 2016 Apr 2];33:559–70. Available from: http://linkinghub.elsevier.com/retrieve/pii/S0272523112000706 142.  Bettelli E, Carrier Y, Gao W, Korn T, Strom TB, Oukka M, et al. Reciprocal developmental pathways for the generation of pathogenic effector TH17 and regulatory T cells. Nature [Internet] 2006 [cited 2016 Apr 2];441:235–8. Available from: http://www.nature.com/doifinder/10.1038/nature04753 143.  Yang J, Chu Y, Yang X, Gao D, Zhu L, Yang X, et al. Th17 and natural Treg cell population dynamics in systemic lupus erythematosus. Arthritis Rheum [Internet] 2009 [cited 2016 Apr 2];60:1472–83. Available from: http://doi.wiley.com/10.1002/art.24499  169 144.  Cua DJ, Kastelein RA. TGF-β, a “double agent” in the immune pathology war. Nat Immunol 2006;7:557–9.  145.  Zhao J, Lloyd CM, Noble A. Th17 responses in chronic allergic airway inflammation abrogate regulatory T-cell-mediated tolerance and contribute to airway remodeling. Mucosal Immunol [Internet] 2013 [cited 2016 Apr 2];6:335–46. Available from: http://www.nature.com/doifinder/10.1038/mi.2012.76 146.  Brusko T, Bluestone J. Clinical application of regulatory T cells for treatment of type 1 diabetes and transplantation. Eur J Immunol [Internet] 2008 [cited 2016 Apr 2];38:931–4. Available from: http://doi.wiley.com/10.1002/eji.200738108 147.  Hoffmann P, Eder R, Boeld TJ, Doser K, Piseshka B, Andreesen R, et al. Only the CD45RA+ subpopulation of CD4+CD25high T cells gives rise to homogeneous regulatory T-cell lines upon in vitro expansion. Blood [Internet] 2006 [cited 2016 Apr 2];108:4260–7. Available from: http://www.bloodjournal.org/cgi/doi/10.1182/blood-2006-06-027409 148.  Nettenstrom L, Alderson K, Raschke EE, Evans MD, Sondel PM, Olek S, et al. An optimized multi-parameter flow cytometry protocol for human T regulatory cell analysis on fresh and viably frozen cells, correlation with epigenetic analysis, and comparison of cord and adult blood. J Immunol Methods [Internet] 2013 [cited 2016 Apr 2];387:81–8. Available from: http://linkinghub.elsevier.com/retrieve/pii/S0022175912002967 149.  Moniuszko M, Kowal K, Zukowski S, Dabrowska M, Bodzenta-Lukaszyk A. Frequencies of circulating CD4+CD25+CD127low cells in atopics are altered by bronchial allergen challenge. Eur J Clin Invest [Internet] 2008 [cited 2016 Apr 2];38:201–4. Available from: http://doi.wiley.com/10.1111/j.1365-2362.2007.01920.x 150.  Bajoriūnienė I, Malakauskas K, Lavinskienė S, Jeroch J, Gasiūnienė E, Vitkauskienė A, et al. Peripheral blood Th17 cells and neutrophils in Dermatophagoides pteronyssinus-induced early-and late-phase asthmatic response. Med Kaunas Lith [Internet] 2011 [cited 2016 Apr 2];48:442–51. Available from: http://europepmc.org/abstract/med/23168918 151.  Zhao Y, Yang J, Gao Y, Guo W. Th17 Immunity in Patients with Allergic Asthma. Int Arch Allergy Immunol [Internet] 2010 [cited 2016 Apr 2];151:297–307. Available from: http://www.karger.com/doi/10.1159/000250438 152.  Vilches C, Parham P. Diverse, Rapidly Evolving Receptors of Innate and Adaptive Immunity. Annu Rev Immunol [Internet] 2002 [cited 2016 Apr 2];20:217–51. Available from: http://www.annualreviews.org/doi/abs/10.1146/annurev.immunol.20.092501.134942 153.  Timonen T, Stenius-Aarniala B. Natural killer cell activity in asthma. Clin Exp Immunol [Internet] 1985 [cited 2016 Apr 2];59:85. Available from: http://www.ncbi.nlm.nih.gov/pmc/articles/PMC1577172/ 154.  Korsgren M, Persson CG, Sundler F, Bjerke T, Hansson T, Chambers BJ, et al. Natural killer cells determine development of allergen-induced eosinophilic airway inflammation in  170 mice. J Exp Med [Internet] 1999 [cited 2016 Apr 2];189:553–62. Available from: http://jem.rupress.org/content/189/3/553.abstract 155.  Hammad H, Lambrecht BN. Dendritic cells and epithelial cells: linking innate and adaptive immunity in asthma. Nat Rev Immunol [Internet] 2008 [cited 2016 Apr 2];8:193–204. Available from: http://www.nature.com/doifinder/10.1038/nri2275 156.  Hatzivlassiou M, Grainge C, Kehagia V, Lau L, Howarth PH. The allergen specificity of the late asthmatic reaction. Allergy [Internet] 2010 [cited 2016 Apr 2];65:355–8. Available from: http://doi.wiley.com/10.1111/j.1398-9995.2009.02184.x 157.  Cockcroft D, Davis B. Mechanisms of airway hyperresponsiveness. J Allergy Clin Immunol [Internet] 2006 [cited 2016 Apr 7];118:551–9. Available from: http://linkinghub.elsevier.com/retrieve/pii/S0091674906015119 158.  Imaoka H, Gauvreau GM, Watson RM, Strinich T, Obminksi GL, Howie K, et al. Sputum inflammatory cells and allergen-induced airway responses in allergic asthmatic subjects: Allergen-induced airway inflammation. Allergy [Internet] 2011 [cited 2016 Apr 7];66:1075–80. Available from: http://doi.wiley.com/10.1111/j.1398-9995.2011.02588.x 159.  Pelikan Z. Effects of inhaled corticosteriods on the dual asthmatic response. Allergy Asthma Proc 2013;34:e47–58.  160.  Guilbert TW, Morgan WJ, Zeiger RS, Mauger DT, Boehmer SJ, Szefler SJ, et al. Long-term inhaled corticosteroids in preschool children at high risk for asthma. N Engl J Med [Internet] 2006 [cited 2016 Apr 7];354:1985–97. Available from: http://www.nejm.org/doi/full/10.1056/NEJMoa051378 161.  Vignola AM, Chanez P, Campbell AM, Souques F, Lebel B, Enander I, et al. Airway inflammation in mild intermittent and in persistent asthma. Am J Respir Crit Care Med [Internet] 1998 [cited 2016 Apr 7];157:403–9. Available from: http://www.atsjournals.org/doi/abs/10.1164/ajrccm.157.2.96-08040 162.  FDA-NIH Biomarker Working Group. BEST (Biomarkers, EndpointsS, and other Tools) Resource [Internet]. Maryland: Food and Drug Administration (US)/National Institutes of Health (US); 2016. Available from: http://www.ncbi.nlm.nih.gov/books/NBK326791/ 163.  An Official ATS. American Thoracic Society Documents. Am J Respir Crit Care Med [Internet] 2011 [cited 2016 Apr 5];184:602–15. Available from: http://www.thoracic.org/newsroom/press-releases/resources/ats-document-final.pdf 164.  Essat M, Harnan S, Gomersall T, Tappenden P, Wong R, Pavord I, et al. Fractional exhaled nitric oxide for the management of asthma in adults: a systematic review. Eur Respir J [Internet] 2016 [cited 2016 Apr 7];47:751–68. Available from: http://erj.ersjournals.com/lookup/doi/10.1183/13993003.01882-2015 165.  Conway SJ, Izuhara K, Kudo Y, Litvin J, Markwald R, Ouyang G, et al. The role of periostin in tissue remodeling across health and disease. Cell Mol Life Sci [Internet] 2014  171 [cited 2016 Mar 29];71:1279–88. Available from: http://link.springer.com/10.1007/s00018-013-1494-y 166.  Arron JR, Izuhara K. Asthma biomarkers: what constitutes a “gold standard”? Thorax [Internet] 2015 [cited 2015 Nov 30];70:105–7. Available from: http://thorax.bmj.com/content/70/2/105.short 167.  Zhang X-Y, Simpson JL, Powell H, Yang IA, Upham JW, Reynolds PN, et al. Full blood count parameters for the detection of asthma inflammatory phenotypes. Clin Exp Allergy [Internet] 2014 [cited 2016 Mar 29];44:1137–45. Available from: http://doi.wiley.com/10.1111/cea.12345 168.  Wagener AH, de Nijs SB, Lutter R, Sousa AR, Weersink EJM, Bel EH, et al. External validation of blood eosinophils, FE NO and serum periostin as surrogates for sputum eosinophils in asthma. Thorax [Internet] 2015 [cited 2016 Mar 29];70:115–20. Available from: http://thorax.bmj.com/lookup/doi/10.1136/thoraxjnl-2014-205634 169.  Li X -j., Hayward C, Fong P-Y, Dominguez M, Hunsucker SW, Lee LW, et al. A Blood-Based Proteomic Classifier for the Molecular Characterization of Pulmonary Nodules. Sci Transl Med [Internet] 2013 [cited 2015 Dec 7];5:207ra142–207ra142. Available from: http://stm.sciencemag.org/cgi/doi/10.1126/scitranslmed.3007013 170.  Whitney DH, Elashoff MR, Porta-Smith K, Gower AC, Vachani A, Ferguson JS, et al. Derivation of a bronchial genomic classifier for lung cancer in a prospective study of patients undergoing diagnostic bronchoscopy. BMC Med Genomics [Internet] 2015 [cited 2015 Dec 2];8. Available from: http://www.biomedcentral.com/1755-8794/8/18 171.  Silvestri GA, Vachani A, Whitney D, Elashoff M, Porta Smith K, Ferguson JS, et al. A Bronchial Genomic Classifier for the Diagnostic Evaluation of Lung Cancer. N Engl J Med [Internet] 2015 [cited 2015 Dec 1];373:243–51. Available from: http://www.nejm.org/doi/abs/10.1056/NEJMoa1504601 172.  Huang H-H, Liu X-Y, Liang Y, Chai H, Xia L-Y. Identification of 13 blood-based gene expression signatures to accurately distinguish tuberculosis from other pulmonary diseases and healthy controls. Biomed Mater Eng [Internet] 2015 [cited 2016 Mar 4];26:S1837–43. Available from: http://www.medra.org/servlet/aliasResolver?alias=iospress&doi=10.3233/BME-151486 173.  Cohen Freue GV, Meredith A, Smith D, Bergman A, Sasaki M, Lam KKY, et al. Computational Biomarker Pipeline from Discovery to Clinical Implementation: Plasma Proteomic Biomarkers for Cardiac Transplantation. PLoS Comput Biol [Internet] 2013 [cited 2016 Feb 15];9:e1002963. Available from: http://dx.plos.org/10.1371/journal.pcbi.1002963 174.  Century I of M (US) C on A the H of the P in the 21st. The Future of the Public’s Health in the 21st Century. National Academy Press; 2003.  175.  Singh A, Yamamoto M, Ruan J, Choi JY, Gauvreau GM, Olek S, et al. Th17/Treg ratio derived using DNA methylation analysis is associated with the late phase asthmatic  172 response. Allergy Asthma Clin Immunol [Internet] 2014 [cited 2016 Mar 2];10:32. Available from: http://www.biomedcentral.com/content/pdf/1710-1492-10-32.pdf 176.  Haas BJ, Papanicolaou A, Yassour M, Grabherr M, Blood PD, Bowden J, et al. De novo transcript sequence reconstruction from RNA-seq using the Trinity platform for reference generation and analysis. Nat Protoc [Internet] 2013 [cited 2016 Mar 2];8:1494–512. Available from: http://www.nature.com.ezproxy.library.ubc.ca/nprot/journal/v8/n8/abs/nprot.2013.084.html 177.  Li B, Dewey CN. RSEM: accurate transcript quantification from RNA-Seq data with or without a reference genome. BMC Bioinformatics [Internet] 2011 [cited 2016 Mar 2];12:1. Available from: https://bmcbioinformatics.biomedcentral.com/articles/10.1186/1471-2105-12-323 178.  Dobin A, Davis CA, Schlesinger F, Drenkow J, Zaleski C, Jha S, et al. STAR: ultrafast universal RNA-seq aligner. Bioinformatics [Internet] 2013 [cited 2016 Mar 3];29:15–21. Available from: https://bioinformatics.oxfordjournals.org/content/29/1/15.full 179.  Harrow J, Frankish A, Gonzalez JM, Tapanari E, Diekhans M, Kokocinski F, et al. GENCODE: The reference human genome annotation for The ENCODE Project. Genome Res [Internet] 2012 [cited 2016 Mar 4];22:1760–74. Available from: http://genome.cshlp.org/cgi/doi/10.1101/gr.135350.111 180.  Liao Y, Smyth GK, Shi W. featureCounts: an efficient general purpose program for assigning sequence reads to genomic features. Bioinformatics [Internet] 2014 [cited 2016 Mar 3];30:923–30. Available from: http://bioinformatics.oxfordjournals.org/cgi/doi/10.1093/bioinformatics/btt656 181.  Law CW, Chen Y, Shi W, Smyth GK. Voom: precision weights unlock linear model analysis tools for RNA-seq read counts. Genome Biol [Internet] 2014 [cited 2016 Mar 2];15:R29. Available from: http://www.biomedcentral.com/content/pdf/gb-2014-15-2-r29.pdf 182.  Vandesompele J, De Preter K, Pattyn F, Poppe B, Van Roy N, De Paepe A, et al. Accurate normalization of real-time quantitative RT-PCR data by geometric averaging of multiple internal control genes. Genome Biol [Internet] 2002 [cited 2016 Mar 6];3:1–12. Available from: http://link.springer.com/article/10.1186/gb-2002-3-7-research0034 183.  NanoString Technologies, Inc. nCounter Expression Data Analysis Guide.  184.  Zou H, Hastie T. Regularization and variable selection via the elastic net. J R Stat Soc Ser B Stat Methodol [Internet] 2005 [cited 2015 Jul 15];67:301–20. Available from: http://onlinelibrary.wiley.com/doi/10.1111/j.1467-9868.2005.00503.x/pdf 185.  Breiman L. Random forests. Mach Learn [Internet] 2001 [cited 2016 Mar 3];45:5–32. Available from: http://link.springer.com/article/10.1023/A:1010933404324  173 186.  Hughey JJ, Butte AJ. Robust meta-analysis of gene expression using the elastic net. Nucleic Acids Res [Internet] 2015 [cited 2015 Aug 14];43:e79–e79. Available from: http://nar.oxfordjournals.org/content/43/12/e79 187.  Van Vliet D, Smolinska A, Jöbsis Q, Rosias PPR, Muris JWM, Dallinga JW, et al. Association between exhaled inflammatory markers and asthma control in children. J Breath Res [Internet] 2016 [cited 2016 Mar 3];10:016014. Available from: http://stacks.iop.org/1752-7163/10/i=1/a=016014?key=crossref.764cedcf36ad2e912939099c4db4489f 188.  Prosperi MC, Marinho S, Simpson A, Custovic A, Buchan IE. Predicting phenotypes of asthma and eczema with machine learning. BMC Med Genomics [Internet] 2014 [cited 2016 Mar 4];7:1. Available from: http://bmcmedgenomics.biomedcentral.com/articles/10.1186/1755-8794-7-S1-S7 189.  Sin DD, Hollander Z, DeMarco ML, McManus BM, Ng RT. Biomarker Development for Chronic Obstructive Pulmonary Disease. From Discovery to Clinical Implementation. Am J Respir Crit Care Med [Internet] 2015 [cited 2016 Feb 28];192:1162–70. Available from: http://www.atsjournals.org/doi/10.1164/rccm.201505-0871PP 190.  Youden WJ. Index for rating diagnositc tests. Cancer 1950;3:32–5.  191.  Chen EY, Tan CM, Kou Y, Duan Q, Wang Z, Meirelles GV, et al. Enrichr: interactive and collaborative HTML5 gene list enrichment analysis tool. BMC Bioinformatics [Internet] 2013 [cited 2016 Mar 4];14:128. Available from: http://www.biomedcentral.com/1471-2105/14/128?utm_source=feedburner&utm_medium=feed&utm_campaign=Feed%3A+bioinformatics_today+(Bioinformatics) 192.  NanoString Technologies, Inc. Strategies for Successful Gene Expression Assays.  193.  Horvatovich P, Bischoff R, Thurston DE. Comprehensive Biomarker Discovery and Validation for Clinical Application. 2013: RSC Publishing;  194.  Baldwin AS. Series Introduction: The transcription factor NF-κB and human disease. J Clin Invest [Internet] 2001 [cited 2016 Apr 7];107:3–6. Available from: http://www.jci.org/articles/view/11891 195.  Covic M, Hassa PO, Saccani S, Buerki C, Meier NI, Lombardi C, et al. Arginine methyltransferase CARM1 is a promoter-specific regulator of NF-κB-dependent gene expression. EMBO J [Internet] 2005 [cited 2016 Apr 7];24:85–96. Available from: http://emboj.embopress.org/content/24/1/85.abstract 196.  Baines KJ, Hsu AC, Tooze M, Gunawardhana LP, Gibson PG, Wark PA. Novel immune genes associated with excessive inflammatory and antiviral responses to rhinovirus in COPD. Respir Res [Internet] 2013 [cited 2016 Apr 8];14:15. Available from: http://www.biomedcentral.com/content/pdf/1465-9921-14-15.pdf  174 197.  Peters VA, Joesting JJ, Freund GG. IL-1 receptor 2 (IL-1R2) and its role in immune regulation. Brain Behav Immun [Internet] 2013 [cited 2016 Apr 8];32:1–8. Available from: http://linkinghub.elsevier.com/retrieve/pii/S0889159112004928 198.  Hernandez ML, Mills K, Almond M, Todoric K, Aleman MM, Zhang H, et al. IL-1 receptor antagonist reduces endotoxin-induced airway inflammation in healthy volunteers. J Allergy Clin Immunol [Internet] 2015 [cited 2016 Apr 7];135:379–85. Available from: http://linkinghub.elsevier.com/retrieve/pii/S0091674914010434 199.  Ame-Thomas P, Hoeller S, Artchounin C, Misiak J, Braza MS, Le Priol J, et al. CD10 delineates a subset of human IL-4 producing follicular helper T cells involved in the survival of follicular lymphoma B cells. Blood 2015;125:2381–5.  200.  Gantke T, Sriskantharajah S, Ley SC. Regulation and function of TPL-2, an IκB kinase-regulated MAP kinase kinase kinase. Cell Res [Internet] 2011 [cited 2016 Apr 8];21:131–45. Available from: http://www.nature.com/cr/journal/v21/n1/abs/cr2010173a.html 201.  Malek S, Huxford T, Ghosh G. IkBa functions through direct contacts with the nuclear localization signals and the DNA binding sequences of NF-kB. J Biol Chem 1998;273:25427–35.  202.  Li Z-W, Chu W, Hu Y, Delhase M, Deerinck T, Ellisman M, et al. The IKKB Subunit of IkB Kinase (IKK) is Essential for Nuclear Factor kB Activation and Prevention of Apoptosis. J Exp Med 1999;189:1839–45.  203.  Thome M. CARMA1, BCL-10 and MALT1 in lymphocyte development and activation. Nat Rev Immunol [Internet] 2004 [cited 2016 Apr 8];4:348–59. Available from: http://www.nature.com/doifinder/10.1038/nri1352 204.  Klemm S, Gutermuth J, Hültner L, Sparwasser T, Behrendt H, Peschel C, et al. The Bcl10–Malt1 complex segregates FcεRI-mediated nuclear factor κB activation and cytokine production from mast cell degranulation. J Exp Med [Internet] 2006 [cited 2016 Apr 8];203:337–47. Available from: http://www.jem.org/lookup/doi/10.1084/jem.20051982 205.  Ashkenazi A, Dixit V. Death Receptors: Signaling and Modulation. Science 1998;281:1305–8.  206.  Qi X, Gurung P, Malireddi RKS, Karmaus PWF, Sharma D, Vogel P, et al. Critical role of caspase-8-mediated IL-1 signaling in promoting Th2 responses during asthma pathogenesis. Mucosal Immunol [Internet] 2016 [cited 2016 Apr 7];Available from: http://www.nature.com/doifinder/10.1038/mi.2016.25 207.  Mansour NM, Bernal GM, Wu L, Crawley CD, Cahill KE, Voce DJ, et al. Decoy Receptor DcR1 Is Induced in a p50/Bcl3-Dependent Manner and Attenuates the Efficacy of Temozolomide. Cancer Res [Internet] 2015 [cited 2016 Apr 8];75:2039–48. Available from: http://cancerres.aacrjournals.org/cgi/doi/10.1158/0008-5472.CAN-14-2144  175 208.  Migeotte I, Communi D, Parmentier M. Formyl peptide receptors: A promiscuous subfamily of G protein-coupled receptors controlling immune responses. Cytokine Growth Factor Rev [Internet] 2006 [cited 2016 Apr 6];17:501–19. Available from: http://linkinghub.elsevier.com/retrieve/pii/S135961010600058X 209.  Fiore S, Maddox JF, Perez HD, Serhan CN. Identification of a human cDNA encoding a functional high affinity lipoxin A4 receptor. J Exp Med [Internet] 1994 [cited 2016 Apr 6];180:253–60. Available from: http://jem.rupress.org/content/180/1/253.abstract 210.  Levy BD, De Sanctis GT, Devchand PR, Kim E, Ackerman K, Schmidt BA, et al. Multi-pronged inhibition of airway hyper-responsiveness and inflammation by lipoxin A4. Nat Med [Internet] 2002 [cited 2016 Apr 6];8:1018–23. Available from: http://www.nature.com/doifinder/10.1038/nm748 211.  Gagliardo R, Gras D, La Grutta S, Chanez P, Di Sano C, Albano GD, et al. Airway lipoxin A4/formyl peptide receptor 2–lipoxin receptor levels in pediatric patients with severe asthma. J Allergy Clin Immunol [Internet] 2016 [cited 2016 Apr 6];Available from: http://linkinghub.elsevier.com/retrieve/pii/S0091674916001949 212.  Perretti M, Chiang N, La M, Fierro IM, Marullo S, Getting SJ, et al. Endogenous lipid- and peptide-derived anti-inflammatory pathways generated with glucocorticoid and aspirin treatment activate the lipoxin A4 receptor. Nat Med [Internet] 2002 [cited 2016 Apr 6];8:1296–302. Available from: http://www.nature.com/doifinder/10.1038/nm786 213.  Ernst S, Lange C, Wilbers A, Goebeler V, Gerke V, Rescher U. An Annexin 1 N-Terminal Peptide Activates Leukocytes by Triggering Different Members of the Formyl Peptide Receptor Family. J Immunol [Internet] 2004 [cited 2016 Apr 6];172:7669–76. Available from: http://www.jimmunol.org/cgi/doi/10.4049/jimmunol.172.12.7669 214.  Perretti M, Dalli J. Exploiting the Annexin A1 pathway for the development of novel anti-inflammatory therapeutics: Annexin A1 and endogenous anti-inflammation. Br J Pharmacol [Internet] 2009 [cited 2016 Apr 9];158:936–46. Available from: http://doi.wiley.com/10.1111/j.1476-5381.2009.00483.x 215.  Errasfa M, Russo-Marie F. A purified lipocortin shares the anti-inflammatory effect of glucocorticosteroids in vivo in mice. Br J Pharmacol [Internet] 1989 [cited 2016 Apr 9];97:1051–8. Available from: http://onlinelibrary.wiley.com/doi/10.1111/j.1476-5381.1989.tb12561.x/abstract 216.  Van Hal P, Overbeek S, Hoogsteden H, Zijlstra F, Murphy K, Oosterhoff Y, et al. Eicosanoids and lipocortin-1 in BAL fluid in asthma: effects of smoking and inhaled glucocorticoids. J Appl Physiol 1996;81:548–55.  217.  Kapranov P, St Laurent G, Raz T, Ozsolak F, Reynolds CP, Sorensen PH, et al. The majority of total nuclear-encoded non-ribosomal RNA in a human cell is’ dark matter’un-annotated RNA. BMC Biol [Internet] 2010 [cited 2016 Apr 9];8:149. Available from: http://www.biomedcentral.com/1741-7007/8/149  176 218.  Whitehead J, Pandey GK, Kanduri C. Regulation of the mammalian epigenome by long noncoding RNAs. Biochim Biophys Acta BBA - Gen Subj [Internet] 2009 [cited 2016 Apr 9];1790:936–47. Available from: http://linkinghub.elsevier.com/retrieve/pii/S0304416508002389 219.  Gupta RA, Shah N, Wang KC, Kim J, Horlings HM, Wong DJ, et al. Long non-coding RNA HOTAIR reprograms chromatin state to promote cancer metastasis. Nature [Internet] 2010 [cited 2016 Apr 9];464:1071–6. Available from: http://www.nature.com/doifinder/10.1038/nature08975 220.  Lee GL, Dobi A, Srivastava S. Prostate cancer: Diagnostic performance of the PCA3 urine test. Nat Rev Urol [Internet] 2011 [cited 2016 Apr 9];8:123–4. Available from: http://www.nature.com/doifinder/10.1038/nrurol.2011.10 221.  Ritchie MD, Holzinger ER, Li R, Pendergrass SA, Kim D. Methods of integrating data to uncover genotype–phenotype interactions. Nat Rev Genet [Internet] 2015 [cited 2015 Jul 10];16:85–97. Available from: http://www.nature.com/doifinder/10.1038/nrg3868 222.  Günther O, Chen V, Freue GC, Balshaw R, Tebbutt S, Hollander Z, et al. A Computational Pipeline for the Development of Multi-Marker Bio-Signature Panels and Ensemble Classifiers. 2012 [cited 2016 Jan 19];13:326. Available from: http://summit.sfu.ca/item/13303 223.  Bailey RC, Kwong GA, Radu CG, Witte ON, Heath JR. DNA-Encoded Antibody Libraries: A Unified Platform for Multiplexed Cell Sorting and Detection of Genes and Proteins. J Am Chem Soc [Internet] 2007 [cited 2016 Mar 27];129:1959–67. Available from: http://pubs.acs.org/doi/abs/10.1021/ja065930i 224.  NanoString Technologies. 3D Biology: Simultaneous multiplexed digital detection of DNA [Internet]. 2016;Available from: www.nanostring.com 225.  Langfelder P, Horvath S. WGCNA: an R package for weighted correlation network analysis. BMC Bioinformatics [Internet] 2008 [cited 2016 Apr 4];9:559. Available from: http://www.biomedcentral.com/1471-2105/9/559 226.  Chikina M, Zaslavsky E, Sealfon SC. CellCODE: a robust latent variable approach to differential expression analysis for heterogeneous cell populations. Bioinformatics [Internet] 2015 [cited 2016 Apr 4];31:1584–91. Available from: http://bioinformatics.oxfordjournals.org/cgi/doi/10.1093/bioinformatics/btv015 227.  Leek JT, Storey JD. Capturing heterogeneity in gene expression studies by surrogate variable analysis. PLoS Genet [Internet] 2007 [cited 2016 Apr 4];3:e161. Available from: http://journals.plos.org/plosgenetics/article?id=10.1371/journal.pgen.0030161 228.  Abbas AR, Wolslegel K, Seshasayee D, Modrusan Z, Clark HF. Deconvolution of Blood Microarray Data Identifies Cellular Activation Patterns in Systemic Lupus Erythematosus. PLoS ONE [Internet] 2009 [cited 2016 Apr 4];4:e6098. Available from: http://dx.plos.org/10.1371/journal.pone.0006098  177 229.  Novershtern N, Subramanian A, Lawton LN, Mak RH, Haining WN, McConkey ME, et al. Densely Interconnected Transcriptional Circuits Control Cell States in Human Hematopoiesis. Cell [Internet] 2011 [cited 2016 Apr 4];144:296–309. Available from: http://linkinghub.elsevier.com/retrieve/pii/S0092867411000055 230.  Newman AM, Liu CL, Green MR, Gentles AJ, Feng W, Xu Y, et al. Robust enumeration of cell subsets from tissue expression profiles. Nat Methods [Internet] 2015 [cited 2016 Apr 4];12:453–7. Available from: http://www.nature.com/doifinder/10.1038/nmeth.3337 231.  Haw R, Hermjakob H, D’Eustachio P, Stein L. Reactome pathway analysis to enrich biological discovery in proteomics data sets. PROTEOMICS [Internet] 2011 [cited 2016 Apr 5];11:3598–613. Available from: http://doi.wiley.com/10.1002/pmic.201100066 232.  Yamamoto S, Sakai N, Nakamura H, Fukagawa H, Fukuda K, Takagi T. INOH: ontology-based highly structured database of signal transduction pathways. Database [Internet] 2011 [cited 2016 Apr 5];2011:bar052–bar052. Available from: http://database.oxfordjournals.org/cgi/doi/10.1093/database/bar052 233.  Kandasamy K, Mohan SS, Raju R, Keerthikumar S, Kumar GSS, Venugopal AK, et al. NetPath: a public resource of curated signal transduction pathways. Genome Biol [Internet] 2010;11:R3–R3. Available from: http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2847715/ 234.  Breuer K, Foroushani AK, Laird MR, Chen C, Sribnaia A, Lo R, et al. InnateDB: systems biology of innate immunity and beyond--recent updates and continuing curation. Nucleic Acids Res [Internet] 2013 [cited 2016 Apr 5];41:D1228–33. Available from: http://nar.oxfordjournals.org/lookup/doi/10.1093/nar/gks1147 235.  Lee HK, Hsu AK, Sajdak J, Qin J, Pavlidis P. Coexpression analysis of human genes across many microarray data sets. Genome Res [Internet] 2004 [cited 2016 Mar 30];14:1085–94. Available from: http://genome.cshlp.org/content/14/6/1085.short 236.  Liquet B, Lê Cao K-A, Hocini H, Thiébaut R. A novel approach for biomarker selection and the integration of repeated measures experiments from two assays. BMC Bioinformatics [Internet] 2012 [cited 2015 Jul 18];13:325. Available from: http://www.biomedcentral.com/1471-2105/13/325/ 237.  Stone KD, Prussin C, Metcalfe DD. IgE, mast cells, basophils, and eosinophils. J Allergy Clin Immunol [Internet] 2010 [cited 2016 Apr 5];125:S73–80. Available from: http://linkinghub.elsevier.com/retrieve/pii/S0091674909017345 238.  Marone G, Triggiani M, Depaulis A. Mast cells and basophils: friends as well as foes in bronchial asthma? Trends Immunol [Internet] 2005 [cited 2016 Apr 5];26:25–31. Available from: http://linkinghub.elsevier.com/retrieve/pii/S1471490604003345 239.  Schroeder JT, Chichester KL, Bieneman AP. Human Basophils Secrete IL-3: Evidence of Autocrine Priming for Phenotypic and Functional Responses in Allergic Disease. J Immunol [Internet] 2009 [cited 2016 Apr 5];182:2432–8. Available from: http://www.jimmunol.org/cgi/doi/10.4049/jimmunol.0801782  178 240.  Kim S, Shen T, Min B. Basophils Can Directly Present or Cross-Present Antigen to CD8 Lymphocytes and Alter CD8 T Cell Differentiation into IL-10-Producing Phenotypes. J Immunol [Internet] 2009 [cited 2016 Apr 5];183:3033–9. Available from: http://www.jimmunol.org/cgi/doi/10.4049/jimmunol.0900332 241.  Cui X, Geffers L, Eichele G, Yan J. A Computational Approach to Estimate Interorgan Metabolic Transport in a Mammal. PLoS ONE [Internet] 2014 [cited 2016 Apr 6];9:e100963. Available from: http://dx.plos.org/10.1371/journal.pone.0100963 242.  Comhair SAA, McDunn J, Bennett C, Fettig J, Erzurum SC, Kalhan SC. Metabolomic Endotype of Asthma. J Immunol [Internet] 2015 [cited 2016 Apr 5];195:643–50. Available from: http://www.jimmunol.org/cgi/doi/10.4049/jimmunol.1500736 243.  Dweik RA, Sorkness RL, Wenzel S, Hammel J, Curran-Everett D, Comhair SAA, et al. Use of Exhaled Nitric Oxide Measurement to Identify a Reactive, at-Risk Phenotype among Patients with Asthma. Am J Respir Crit Care Med [Internet] 2010 [cited 2016 Apr 5];181:1033–41. Available from: http://www.atsjournals.org/doi/abs/10.1164/rccm.200905-0695OC 244.  Ugajin T, Satoh T, Kanamori T, Aritake K, Urade Y, Yokozeki H. FcεRI, but Not FcγR, Signals Induce Prostaglandin D2 and E2 Production from Basophils. Am J Pathol [Internet] 2011 [cited 2016 Apr 5];179:775–82. Available from: http://linkinghub.elsevier.com/retrieve/pii/S0002944011004299 245.  Luster AD, Tager AM. T-cell trafficking in asthma: lipid mediators grease the way. Nat Rev Immunol [Internet] 2004 [cited 2016 Apr 5];4:711–24. Available from: http://www.nature.com/doifinder/10.1038/nri1438 246.  Ackerman SJ, Kwatia MA, Doyle CB, Enhorning G. Hydrolysis of surfactant phospholipids catalyzed by phospholipase a2 and eosinophil lysophospholipases causes surfactant dysfunction*: A mechanism for small airway closure in asthma. Chest [Internet] 2003 [cited 2015 Oct 5];123:355S – 355S. Available from: http://dx.doi.org/10.1378/chest.123.3_suppl.355S 247.  Ried JS, Baurecht H, Stückler F, Krumsiek J, Gieger C, Heinrich J, et al. Integrative genetic and metabolite profiling analysis suggests altered phosphatidylcholine metabolism in asthma. Allergy [Internet] 2013 [cited 2015 Jul 11];68:629–36. Available from: http://doi.wiley.com/10.1111/all.12110 248.  Hotze M, Baurecht H, Rodríguez E, Chapman-Rothe N, Ollert M, Fölster-Holst R, et al. Increased efficacy of omalizumab in atopic dermatitis patients with wild-type filaggrin status and higher serum levels of phosphatidylcholines. Allergy [Internet] 2014 [cited 2015 Jul 11];69:132–5. Available from: http://doi.wiley.com/10.1111/all.12234 249.  Rabinovitch N, Graber NJ, Chinchilli VM, Sorkness CA, Zeiger RS, Strunk RC, et al. Urinary leukotriene E4/exhaled nitric oxide ratio and montelukast response in childhood asthma. J Allergy Clin Immunol [Internet] 2010 [cited 2016 Apr 6];126:545–51.e4. Available from: http://linkinghub.elsevier.com/retrieve/pii/S0091674910010560  179 250.  Ono E, Dutile S, Kazani S, Wechsler ME, Yang J, Hammock BD, et al. Lipoxin Generation Is Related to Soluble Epoxide Hydrolase Activity in Severe Asthma. Am J Respir Crit Care Med [Internet] 2014 [cited 2016 Apr 6];190:886–97. Available from: http://www.atsjournals.org/doi/abs/10.1164/rccm.201403-0544OC 251.  Abdallah Chouarfia, Hafida Bouziane, Belhadri Messabih. Profiles and Majority Voting-Based Ensemble Method for Protein Secondary Structure Prediction. Evol Bioinforma [Internet] 2011 [cited 2016 Apr 6];171. Available from: http://www.la-press.com/profiles-and-majority-voting-based-ensemble-method-for-protein-seconda-article-a2870 252.  Davies SK, Ang JE, Revell VL, Holmes B, Mann A, Robertson FP, et al. Effect of sleep deprivation on the human metabolome. Proc Natl Acad Sci [Internet] 2014 [cited 2015 Oct 5];111:10761–6. Available from: http://www.pnas.org/content/111/29/10761 253.  Floegel A, Wientzek A, Bachlechner U, Jacobs S, Drogan D, Prehn C, et al. Linking diet, physical activity, cardiorespiratory fitness and obesity to serum metabolite networks: findings from a population-based study. Int J Obes [Internet] 2014 [cited 2015 Oct 5];38:1388–96. Available from: http://www.nature.com/ijo/journal/v38/n11/full/ijo201439a.html 254.  Kam SHY, Singh A, He J-Q, Ruan J, Gauvreau GM, O’Byrne PM, et al. Peripheral Blood Gene Expression Changes during Allergen Inhalation Challenge in Atopic Asthmatic Individuals. J Asthma [Internet] 2012 [cited 2016 Mar 2];49:219–26. Available from: http://www.tandfonline.com/doi/full/10.3109/02770903.2011.654300 255.  Yamamoto M, Singh A, Ruan J, Gauvreau GM, O’Byrne PM, Carlsten CR, et al. Decreased miR-192 expression in peripheral blood of asthmatic individuals undergoing an allergen inhalation challenge. BMC Genomics [Internet] 2012 [cited 2016 Mar 2];13:1. Available from: http://www.biomedcentral.com/1471-2164/13/655?utm_source=dlvr.it&utm_medium=linkedin 256.  Singh A, Cohen Freue GV, Oosthuizen JL, Kam SHY, Ruan J, Takhar MK, et al. Plasma proteomics can discriminate isolated early from dual responses in asthmatic individuals undergoing an allergen inhalation challenge. PROTEOMICS - Clin Appl [Internet] 2012 [cited 2016 Mar 2];6:476–85. Available from: http://doi.wiley.com/10.1002/prca.201200013 257.  Singh A, Yamamoto M, Kam SHY, Ruan J, Gauvreau GM, O’Byrne PM, et al. Gene-Metabolite Expression in Blood Can Discriminate Allergen-Induced Isolated Early from Dual Asthmatic Responses. PLoS ONE [Internet] 2013 [cited 2015 Jul 18];8:e67907. Available from: http://dx.plos.org/10.1371/journal.pone.0067907 258.  Hankin CS, Bronstone A, Wang Z, Small MB, Buck P. Estimated Prevalence and Economic Burden of Severe, Uncontrolled Asthma in the United States. J Allergy Clin Immunol 2013;131:AB126.  259.  Fazio S. Mild Asthma [Internet]. Mild Asthma2013;Available from: http://blogs.nejm.org/now/index.php/mild-asthma/2013/08/09/  180 260.  Gauvreau GM, O’Byrne PM, Boulet L-P, Wang Y, Cockcroft D, Bigler J, et al. Effects of an anti-TSLP antibody on allergen-induced asthmatic responses. N Engl J Med [Internet] 2014 [cited 2015 Oct 29];370:2102–10. Available from: http://www.nejm.org.ezproxy.library.ubc.ca/doi/full/10.1056/NEJMoa1402895 261.  Laitinen L, Heino M, Laitinen A, Kava T, Haahtela T. Damage of the Airway Epithelium and Bronchial Reactivity in Patients with Asthma. Am Rev Respir Dis 1985;131:599–606.  262.  Tenenhaus A, Philippe C, Guillemot V, Le Cao K-A, Grill J, Frouin V. Variable selection for generalized canonical correlation analysis. Biostatistics [Internet] 2014 [cited 2015 Jul 15];15:569–83. Available from: http://biostatistics.oxfordjournals.org/cgi/doi/10.1093/biostatistics/kxu001 263.  Tenenhaus M. La régression PLS: théorie et pratique. Editions TECHNIP; 1998.    181 Appendices Appendix A  Supplementary material for Chapter 2 A.1! Determining cell-specific miRNA expression using multiple regression The following test statistic was used in the cell-specific Significance Analysis of Microarrays (csSAM) in order to identify cell-specific differentially expressed genes between two groups92.  1.! Perform multiple regression of miRNA expression onto the relative cell-type frequencies for each group g.    2.! The test statistic to determine significant changes in cell-specific miRNA expression. The following test statistic (similar to the Wald test) was used to determine significant miRNA expression changes in the kth cell between two groups. Test statistic comparing group 1 and group 2 for the kth cell-type is as follows:   3.! The p-value for each test-statistic was calculated by generating an empirical distribution by re-calculating test-statistics after reshuffling of class labels 1000 times. A p-value of 0.05 was deemed significant.   yg = βo + β1gx1g +!+ βkgxkg + εgβo = 0; implies that at zero cell-type frequency there is zero miRNA expressionβkg; increase in miRNA expression for 1% increase in the kth  cell-type frequencyin group g (mean miRNA expression in the kth  cell-type in group g)yg; vector of miRNA expression values for group gxkg; vector of relative cell-type frequencies for the kth  cell-type for group gtk21 =^β k2 −^β k1se^β k⎛⎝⎜⎞⎠⎟ 21se^β k⎛⎝⎜⎞⎠⎟ 21=n1 se^β k1⎛⎝⎜⎞⎠⎟⎡⎣⎢⎤⎦⎥2+ n2 se^β k2⎛⎝⎜⎞⎠⎟⎡⎣⎢⎤⎦⎥2n1 + n2 182  A.2! Partial slopes for both granulocytes and PBMCs in HC and asthmatic individuals (pre- and post-challenge).    183 A.3! Optimal number of proteins and components to select for sPLS-DA  The average error rate across 50 iterations of a 10-fold cross-validation for varying number of proteins selected and components using sPLS-DA.    184 Appendix B  Supplementary material for Chapter 3 B.1! Differentially expressed genes and metabolites at post-challenge (normalized to pre-challenge levels, ratio of post to pre levels). A.! Differentially expressed genes at post-challenge (BH-FDR<10%).   FC* in ERs FC in DRs P -Value BH-FDR 8011680 ALOX15 -1.034±0.053 -1.007±0.013 1.37E-06 0.000681 8085062 IL5RA -1.072±0.008 -1.022±0.005 5.80E-06 0.001 8124527 HIST1H1B -1.008±0.009 1.026±0.016 1.25E-05 0.002 7916862 WLS 1.011±0.011 1.036±0.018 1.38E-05 0.002 8047443 STRADB -1.021±0.007 1.011±0.005 0.000193 0.019 7940565 FADS2 -1.032±0.013 1.015±0.008 0.000254 0.020 7995697 LPCAT2 -1.009±0.005 1.015±0.011 0.00028 0.020 8161288 CNTNAP3 1.017±0.012 1.058±0.024 0.000422 0.026 7953901 CLEC12A -1.030±0.015 1.018±0.018 0.000609 0.034 8166632 GK 1.007±0.011 1.029±0.018 0.00087 0.043 8174361 TSC22D3 -1.037±0.006 -1.014±0.011 0.001 0.046 8169365 TMEM164 -1.004±0.006 1.016±0.007 0.001 0.046 8176709 CYorf15B -1.003±0.018 1.028±0.038 0.001 0.050 8007607 RUNDC3A -1.021±0.005 1.014±0.019 0.001 0.050 8161460 CNTNAP3 1.037±0.015 1.081±0.032 0.002 0.058 8025103 EMR1 -1.040±0.005 -1.007±0.009 0.002 0.063 7988033 EPB42 -1.028±0.008 1.015±0.010 0.002 0.063 8164596 C9orf78 -1.016±0.007 1.003±0.002 0.002 0.063 7921873 FCGR3A 1.011±0.007 1.020±0.011 0.003 0.084 8153652 SHARPIN -1.024±0.008 1.009±0.007 0.004 0.087 8135378 PRKAR2B -1.015±0.008 1.021±0.005 0.004 0.087 7961230 CSDA -1.027±0.007 1.008±0.009 0.004 0.093 7929052 IFIT3 1.000±0.01 1.033±0.023 0.004 0.093 8171834 RPL9 1.022±0.02 -1.029±0.010 0.005 0.093 8157264 SLC31A2 -1.019±0.005 1.009±0.012 0.005 0.093 *Levels at post-challenge are scaled to pre-challenge levels;  FC = post/pre if FC>0, FC = -1/(post/pre) Values expressed as Mean±SE (standard error)    185  B.! Differentially expressed metabolites at post-challenge (p-value<0.05). BIOCHEMICAL SUB_PATHWAY HMDB FC* in ERs* FC in DRs P -Value andro steroid monosulfate 1* Sterol/Steroid HMDB02759 1.003±0.120 1.054±0.202 0.002 4-hydroxyphenylacetate Phenylalanine & tyrosine metabolism HMDB00020 -1.519±0.210 1.226±0.414 0.006 thymol sulfate Food component/Plant HMDB01878 -1.085±0.146 -1.381±0.317 0.014 1-pentadecanoylglycerophosphocholine* Lysolipid  1.404±0.569 -1.005±0.300 0.014 2-arachidonoylglycerophosphocholine* Lysolipid  3.950±2.892 1.181±0.436 0.014 cortisol Sterol/Steroid HMDB00063 -1.560±0.210 1.008±0.400 0.017 3-methyl-2-oxobutyrate Valine, leucine and isoleucine metabolism HMDB00019 -1.048±0.407 1.307±0.341 0.018 1-linoleoylglycerophosphocholine Lysolipid  1.046±0.192 1.185±0.255 0.019 7-alpha-hydroxy-3-oxo-4-cholestenoate (7-Hoca) Sterol/Steroid HMDB12458 1.188±0.207 1.416±0.253 0.031 cysteine Cysteine, methionine, SAM, taurine metabolism HMDB00574 -1.037±0.215 -1.404±0.110 0.041 bradykinin, hydroxy-pro(3) Polypeptide HMDB11728 -1.344±0.241 2.788±3.447 0.043 *Levels at post-challenge are scaled to pre-challenge levels;  FC = post/pre if FC>0, FC = -1/(post/pre) Values expressed as Mean±SE (standard error)   186 B.2! Correlation of lymphocyte frequencies obtained using a hematolyzer and DNA methylation analysis.  The lymphocyte frequencies obtained using a hematolyzer and estimated lymphocyte frequencies (sum of T and B cell frequencies) using DNA methylation analysis were significantly (p<0.001) correlated (Spearman). However, the frequency estimates, obtained using DNA methylation analysis were over-estimated for majority of the samples (most samples above the unit line).      187 Appendix C  Supplementary material for Chapter 4 C.1! Phenotypic classification of flippers Discovery  Validation   Criteria for dual responders: Participants that demonstrated a maximum drop in FEV1 greater than 15% between 3-7 hours after allergen challenge (late asthmatic response, LAR) were ●●●2011−03−232011−04−272012−10−02ERsDRs●●●2009−08−122009−10−272010−01−13ERsDRs●●●2009−10−222011−01−112011−01−25ERsDRsFlipper1 Flipper2 Flipper3−30−20−100100 10 20 30 0 10 20 30 0 10 20 30Allergen induced shift (pre PC20 /post PC20 )Maximum drop in FEV 1 (3h−7h)Allergen_cleanLabel●a●a●aFungusHDMRagweedFlipper4 Flipper5 Flipper6 Flipper7●●2009−09−172009−09−24ERsDRs●●2011−03−232011−04−12ERsDRs●●2009−12−212010−02−11ERsDRs●●2010−03−162010−02−11ERsDRs−40−30−20−1000 10 20 30 0 10 20 30 0 10 20 30 0 10 20 30Allergen induced shift (pre PC20 /post PC20 )Maximum drop in FEV 1 (3h−7h)Allergen_cleanLabel●a●a●aCatGrassHDM 188 classified as dual responders (DRs). Subjects that did not achieve this cut-off but have a drop greater than 10% and an allergen induced shift (Pre methacholine PC20/Post methacholine PC20) greater than or equal to 2 were also classified as DRs.  Of subjects challenged multiple times (different days, see date) some displayed variability with respect to their phenotypic characterization. However, given that subjects were able to elicit the late response on any given occasion, they were all classified as dual responders. Factors such as allergen dosage, allergen type, FEV1 measurement and time of year may contribute to the biological variability seen with respect to the phenotypic characterization of subjects.  C.2! RNA-Sequencing alignment results  UCSC genes UCSC gene-isoforms Ensembl transcripts Trinity contigs Trimming seqtk – remove first 5 and last 25 bases seqtk – remove first 5 and last 25 bases seqtk – remove first 5 and last 25 bases seqtk – remove first 12 bases sickle Aligner/Counter Bowite2/RSEM Bowite2/RSEM STAR/subread Bowtie/RSEM Number of transcripts 42,465 89,357 60,155 258,403 Total number of read pairs (Millions) Mean±SD  (Min, Max) 25.2±2.6 (18.0, 31.6) 21.3±2.5 (16.3, 26.5) 24.6±2.5 (17.7, 30.8) Alignment rate (%) 35.6±5.1 (14.8, 48.4)* 83.9±4.3 (66.4, 89.9)** 38.2±4.9 (23.2, 49.8)* 44.9±5.35* (14.80, 55.46) *Alignment to the transcriptome **Alignment to the genome  189 C.3! Establishing lower limit of detection in RNA-Seq data   C.4! NanoString quality control criteria 1.! Imaging QC: % FOV (field of view) must be greater than 75 FOV. 2.! Binding Density QC: Binding density must be between 0.05 and 2.25. 3.! Positive Control Linearity QC: The R2 must be greater than 0.9 between the counts and concentrations of the 6 positive controls. 4.! Positive Control Limit of Detection QC: The second lowest positive control spike in (0.5fM) must have counts greater than the Mean±2SD of the negative controls for each sample. ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● ●●●●●● ●●●●●●●●●● ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● ●●●● ●●●●●●●●● ●●●●●●● ●●●●●●●● ●●●●● ●●●●● ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● ●●●● ●●●●●●●●●● ●●●● ● ●● ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●cutoff = 3−50510151e−01 1e+01 1e+03Concentration (Attomoles/uL)log2 cpm 190 C.5! Principal Component Analysis of all RNA-Seq datasets combined      ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●−1000100−200 −100 0 100PC1 (28.3% explained var.)PC2 (20.6% explained var.)AIC_Date●●●●2009201020112012●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●−1000100−200 −100 0 100PC1 (28.3% explained var.)PC2 (20.6% explained var.)Sex●●FM●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●−1000100−200 −100 0 100PC1 (28.3% explained var.)PC2 (20.6% explained var.)Site●●●LAVALMACUBC●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●−1000100−200 −100 0 100PC1 (28.3% explained var.)PC2 (20.6% explained var.)Allergen●●●●●●CatFungusGrassHDMHorseRagweed 191 C.6! Overlap between biomarker candidates across all RNA-Seq datasets    17018131402400 140124UCSC genes UCSC gene−isoformsEnsembl Trinity 192  C.7! NanoString probe sequences   Position Target Sequence Tm CP Tm RP Target ID Datasets House-keepers 1 ARPC4 971-1070 GACTTTGTGATCCACTTCATGGAGGAGATTGACAAGGAGATCAGTGAGATGAAGCTGTCAGTCAATGCCCGTGCCCGCATTGTGGCTGAAGAGTTCCTTA 83 84 uc003bsz.2,uc003bta.2,uc003btb.2,uc003btc.2,uc003btd.4,uc003btf.4,uc003btg.4,uc003bth.4,uc003bti.4,uc003btj.4,uc003btk.3,uc010hco.1,uc011atj.2,uc021wsu.1 UCSCgenes 2 TMBIM6 1221-1320 GGCCCTTCCTTCCTCATTGTTGTTTGGTATGCGCACAGTTCCTGTGGGACTGGGCCGTGAGTTTTCCATTGGAAAGAAGTTCAGTGGTCCCATTGTTAAC 80 78 uc001rux.2,uc001ruy.2,uc001ruz.2,uc010sml.1 UCSCgenes 3 RHOA 1231-1330 GGTACTCTGGTGAGTCACCACTTCAGGGCTTTACTCCGTAACAGATTTTGTTGGCATAGCTCTGGGGTGGGCAGTTTTTTGAAAATGGGCTCAACCAGAA 79 84 uc003cwu.3,uc010hku.3 UCSCgenes 4 MED13 1386-1485 GCTGGACAACAAGGACAGGCACCATCTTTAGGTCAGCAACAACAAATACTTCCTAAGCACAAGACCAATGAGAAGCAAGAAAAGAGTGAAGAGCCACAGA 80 79 ENSG00000108510 Ensembl 5 TOR1AIP2 206-305 AGGCGTGAGCCACCACGCCTGGCCTCACTCAAACAGTAAACGAGTCACTGTATATATGCCTTATTTCCTGGGTCCATGGGAAGCATGAGTCTGTTGGGAC 84 82 ENSG00000169905 Ensembl 6 WAC 756-855 CCTCTGGACTGAACCCCACATCTGCACCTCCAACATCTGCTTCAGCGGTCCCTGTTTCTCCTGTTCCACAGTCGCCAATACCTCCCTTACTTCAGGACCC 82 80 ENSG00000095787 Ensembl Biomarker candidates 1 ABHD5 956-1055 TAAAATGCACCCTGACATTCCAGTTTCAGTGATCTTTGGCGCCCGATCCTGCATAGATGGCAATTCTGGCACCAGCATCCAGTCCTTACGACCACATTCA 80 84 uc003cmx.3;uc003cmx.3 UCSCgenes UCSCgeneIso 2 AHCTF1 406-505 TAATCATGGAGGAGCCAGTGCAAGCACTCAGCATTTACATCCAAGTCTGCGATGGCTTTTTGGAGTGGCAGCTGTGGTCACTGATGTTGGACAGATCCTT 82 83 uc001ibv.3,uc001ibw.1,uc009xgs.2;ENSG00000153207 UCSCgenes ensembl 3 ATP11A 456-555 ATGCCATGAACCAGTGTCCTGTTCATTTCATTCAGCACGGCAAGCTCGTTCGGAAACAAAGTCGAAAGCTGCG 79 83 uc001vsi.4,uc001vsj.4,uc001vsm.1,uc010ago.3,uc02 UCSCgenes  193   Position Target Sequence Tm CP Tm RP Target ID Datasets AGTTGGGGACATTGTCATGGTTAAGGA 1rmp.1 4 ATP8A1 871-970 GAGGATTTCTGGCAGAATTGAGTGTGAAAGTCCAAACAGACATCTCTACGATTTTGTTGGAAACATAAGGCTTGATGGACATGGCACCGTTCCACTGGGA 81 82 uc003gwq.2,uc003gwr.2,uc003gws.2,uc011byz.1;uc003gws.2;ENSG00000124406 UCSCgenes UCSCgeneIso ensembl 5 B3GNT5 131-230 GGAAGGAAAGCCGACCTCCGATTTGGACATTTAAAGAGCTGGGCTTGAACTTCGTGAGTTTCGCTCTAAACTGCCCTTGAAATGAAGCTGGACTTGGAGG 79 76 uc003flk.3,uc003fll.3,uc003flm.3 UCSCgenes 6 C9orf78 911-1010 AGAAGTTCAAGAAAATGAATAGGCGGTACTGAGTTGTGCAGAGTGGGATGTAAATATCGCCTTCCTCTCCCTATATCCCTCCCATGAAAAATGGCTTCCT 84 79 uc004byo.3,uc004byp.3,uc004byq.1;ENSG00000136819 UCSCgenes ensembl 7 CARM1 1166-1265 TCCTGATGGCCAAGTCTGTCAAGTACACGGTGAACTTCTTAGAAGCCAAAGAAGGAGATTTGCACAGGATAGAAATCCCATTCAAATTCCACATGCTGCA 84 81 uc002mpz.3,uc002mqa.3,uc010dxn.3;uc002mpz.3;ENSG00000142453 UCSCgenes UCSCgeneIso ensembl 8 CASP8 302-401 AGATGGACTTCAGCAGAAATCTTTATGATATTGGGGAACAACTGGACAGTGAAGATCTGGCCTCCCTCAAGTTCCTGAGCCTGGACTACATTCCGCAAAG 82 85 comp55716_c0_seq31 trinity 9 CD300LB 1531-1630 ACGAAAACCATCGCAGGAAATGGCACCCTCCCTTTTCGGTGATGTTGAAATCATGTTACTAATGAAAACTGTCCTAGGGAAGTGGTTCTGTCTCCTCACA 82 80 uc002jkx.3,uc010wqz.1;ENSG00000178789 UCSCgenes ensembl 10 CD4 976-1075 TGGCAGGCGGAGAGGGCTTCCTCCTCCAAGTCTTGGATCACCTTTGACCTGAAGAACAAGGAAGTGTCTGTAAAACGGGTTACCCAGGACCCTAAGCTCC 86 86 uc001qqv.2,uc009zez.2,uc009zfa.2,uc009zfb.2,uc009zfc.2,uc010sfj.2,uc010sfk.2,uc010sfl.2,uc010sfm.1;ENSG00000010610 UCSCgenes ensembl 11 CD4_isoform 308-407 CGCTCCTCCCAGCAGCCACTCAGGGAAAGAAAGTGGTGCTGGGCAAAAAAGGGGATACAGTGGAACTGACCTGTACAGCTTCCCAGAAGAAGAGCATACA 85 86 uc001qqv.2 UCSCgeneIso 12 CD59 731-830 GACTTGAACTAGATTGCATGCTTCCTCCTTTGCTCTTGGGAAGACCAGCTTTGCAGTGACAGCTTGAGTGGGTTCTCTGCAGCCCTCAGATTATTTTTCC 81 78 uc001muv.4 UCSCgeneIso 13 CD8A 1321-1420 GCTCAGGGCTCTTTCCTCCACACCATTCAGGTCTTTCTTTCCGAGGCCCCTGTCTCAGGGTGAGGTGCTTGAGTCTCCAACGGCAAGGGAACAAGTACTT 83 83 uc002srt.3,uc002sru.3,uc002srv.3,uc010ytn.2;uc010ytn.2;ENSG00000153563 UCSCgenes UCSCgeneIso ensembl 14 CDK5RAP3 1403-1502 GCTGGAGGATCTGATTGGCAAGCTTACCAGTCTTCAGCTGCAACACCTGTTTATGATCCTGGCCTCACCAAGGTATGTGGACCGAGTGACTGAATTCCTC 83 83 uc002imq.2,uc002imr.4,uc002ims.4,uc010wlc.3,uc031rda.1;ENSG00000108465 UCSCgenes ensembl  194   Position Target Sequence Tm CP Tm RP Target ID Datasets 15 CECR1 1027-1126 TAGCCAGCCCTCTACAAGCTGTCTTCTTGCACACGCTGTCACTTCCTCTCACTCGTTCTTGAATCAGCTCCATGTGCCCATGAAATCAATGGCCTCTGTA 83 81 comp55863_c0_seq6 trinity 16 CHP1 175-274 CCCTCCTTCCCTCCTGTCGCCGTCTCTTCTGGCGCCGCTGCTCCCGGAGGAGCTCCCGGCACGGCGATGGGTTCTCGGGCCTCCACGTTACTGCGGGACG 87 87 ENSG00000187446 ensembl 17 CISH 559-658 CACCAATGTACGCATTGAGTATGCCGACTCCAGCTTCCGTCTGGACTCCAACTGCTTGTCCAGGCCACGCATCCTGGCCTTTCCGGATGTGGTCAGCCTT 84 86 ENSG00000114737 ensembl 18 CLEC4E 571-670 GAGTTTTTTATTGGACTGTCAGACCAGGTTGTCGAGGGTCAGTGGCAATGGGTGGACGGCACACCTTTGACAAAGTCTCTGAGCTTCTGGGATGTAGGGG 78 85 uc001quo.1;uc001quo.1;ENSG00000166523 UCSCgenes UCSCgeneIso ensembl 19 CNTNAP3 1433-1532 TCGAACGTGGAACAGAGCAGGACATTTGCTTTTCGGCGAACTTCGACGTGGTTCAGGGAGTTTCGTCCTCTTTCTTAAGGATGGCAAGCTCAAACTGAGT 81 79 uc004abi.3,uc004abj.3,uc004abk.1,uc011lqr.2,uc011lqs.1;ENSG00000106714 UCSCgenes ensembl 20 CNTNAP3_isoform 4014-4113 ACATAGTTATTAAAATGGGAATAAGTAAGAAAATAGACCTGAGTCACCACAGAGGAAGTAAATTACACATTGTCATCGGCATTGGAAGGAAAATATACTG 79 78 uc004abk.1 UCSCgeneIso 21 COPB1_isoform 193-292 AAGATGCGGAAGGGGAGCGACTAGGCCGCTTGCGTCTGGGCCTGGCAGAAGGGACCGGATTTTCTGGCATCCTTAAATCTTGTGTCAAGGATTGGTTATA 96 75 uc001mli.2 UCSCgeneIso 22 CTDSP2 1641-1740 GGGGAGAAGCTGAAAGACCAAGACTCTTCCCAAGTTAGCTTGTCTCCTCTCCTGTCACCCTAAGAGCCACTGAGTTGTGTAGGGATGAAGACTATTGAAG 81 82 ENSG00000175215 ensembl 23 CTSA 1541-1640 TGCCACAATGGGACATGTGCAACTTTCTGGTAAACTTACAGTACCGCCGTCTCTACCGAAGCATGAACTCCCAGTATCTGAAGCTGCTTAGCTCACAGAA 79 81 uc002xqh.3,uc002xqi.3,uc002xqj.4,uc010zxi.2;uc002xqh.3;ENSG00000064601 UCSCgenes UCSCgeneIso ensembl 24 CYTB_comp57541_c0_seq1 86-185 ACAAACTTACTATCCGCCATCCCATACATTGGGACAGACCTAGTTCAATGAATCTGAGGAGGCTACTCAGTAGACAGTCCCACCCTCACACGATTCTTTA 81 84 comp57541_c0_seq1 trinity 25 DAP 1791-1890 CTGAGGGAGCATGGCACAGCCTCACACTTGAAAGACGGTGTTTGGTTTCCCATCTAATCAACTTAAGGGAAGCCGGCATGTACCCTTCAAGGCCCTGTCA 82 79 uc003jez.4,uc011cmw.2;ENSG00000112977 UCSCgenes ensembl 26 DCAF6 2196-2295 TCAGATAAGTTCACAGCCAAGCCATTGGATTCCAACTCAGGAGAAAGAAATGACCTCAATCTTGATCGCTCTTGTGGGGTTCCAGAAGAATCTGCTTCAT 83 80 ENSG00000143164 ensembl 27 DESI1 2881-2980 CAAGACCCACTGATTTGCCAGTGTGCATGGAAATAATAGATTAGAGCAGAAACTAGCAGGGACTGTTGTATA 82 83 ENSG00000100418 ensembl  195   Position Target Sequence Tm CP Tm RP Target ID Datasets ATCGTGATCTACTAGCAGAATTGGGCCC 28 F13A1 3197-3296 TTCAGGTCCCCTTTCAGAGATATAATAAGCCCAACAAGTTGAAGAAGCTGGCGGATCTAGTGACCAGATATATAGAAGGACTGCAGCCACTGATTCTCTC 80 85 uc003mwv.3,uc011dib.2;uc003mwv.3;ENSG00000124491 UCSCgenes UCSCgeneIso ensembl 29 FAM8A1 1792-1891 CCTGGGGAAATTGTCTTTGGTGTTTAGAGGAGGGAATGAGAACACAAATTGGATAATCCACTGTCTCCCATCCCAGGAGGTGGTGAGTTGGCTACAAGAG 80 84 uc003ncc.3;uc003ncc.3;ENSG00000137414 UCSCgenes UCSCgeneIso ensembl 30 FNIP1 1265-1364 ACAATTTGTAATCTTTACACGATGCCACGAATTGGAGAACCTGTCTGGCTTACAATGATGTCGGGGACTCCAGAAAAGAACCACCTTTGCTATCGTTTCA 79 80 comp56357_c0_seq8 trinity 31 FPR1_intron_comp17070_c0_seq1 1-100 GGATGTTTGGTTTGAAGCTTTCAGGAGAAAGCAAAAGAGCCCTAACGACTTTATGATGGGTCATGGGGAAATGAGTGTAATACAGAAGCAGTCACCTTTC 82 81 comp17070_c0_seq1 trinity 32 FPR2 1201-1300 GATGGGGTCAGGGATATTTTGAGTTCTGTTCATCCTACCCTAATGCCAGTTCCAGCTTCATCTACCCTTGAGTCATATTGAGGCATTCAAGGATGCACAG 78 80 uc002pxs.4;comp55353_c2_seq11 UCSCgeneIso trinity 33 FUT7 1711-1810 ACTGGCATGAATGAGAGCCGATACCAACGCTTCTTTGCCTGGCGTGACAGGCTCCGCGTGCGACTGTTCACCGACTGGCGGGAACGTTTCTGTGCCATCT 83 84 uc004ckq.2;uc004ckq.2 UCSCgenes UCSCgeneIso 34 GBE1 1051-1150 CTTTGCAGCTTCCAGCCGTTATGGAACACCTGAAGAGCTACAAGAACTGGTAGACACAGCTCATTCCATGGGTATCATAGTCCTCTTAGATGTGGTACAC 81 81 ENSG00000114480 ensembl 35 GNLY 306-405 CAGGAGCTGGGCCGTGACTACAGGACCTGTCTGACGATAGTCCAAAAACTGAAGAAGATGGTGGATAAGCCCACCCAGAGAAGTGTTTCCAATGCTGCGA 82 81 uc002sql.4,uc010fgp.3,uc010ysx.2;ENSG00000115523 UCSCgenes ensembl 36 GTF2H2 896-995 ATCATGTTAGTCCTCCTCCTGCTAGCTCAAGTTCTGAATGCTCACTTATTCGTATGGGATTTCCTCAGCACACCATTGCTTCTTTATCTGACCAGGATGC 74 79 uc003kau.5,uc003kav.5,uc003kay.1,uc003kaz.4,uc011crt.3,uc032uxt.1,uc032uxu.1 UCSCgenes 37 HBA2 437-536 CCTCCCTGGACAAGTTCCTGGCTTCTGTGAGCACCGTGCTGACCTCCAAATACCGTTAAGCTGGAGCCTCGGTAGCCGTTCCTCCTGCCCGCTGGGCCTC 89 90 uc002cfv.4;uc002cfv.4;ENSG00000188536;comp56950_c0_seq1 UCSCgenes UCSCgeneIso Ensemble trinity 38 HCLS1 516-615 GCAGACAAGTCAGCAGTCGGCTTTGATTATAAAGGAGAAGTGGAGAAGCACACATCTCAGAAAGATTACTCTCGTGGCTTTGGTGGCCGGTACGGGGTGG 83 83 comp54057_c0_seq6 trinity 39 HIP1 5006-5105 CTAAGTGGGACATTCAAAAAACTCTCTCCCAGGACC 79 82 ENSG00000127946 ensembl  196   Position Target Sequence Tm CP Tm RP Target ID Datasets TCGGATGACCATACTCAGACGTGTGACCTCCATACTGGGCTAAGGAAGTATCAGCACTAGAAAT 40 HLA-G 1181-1280 AAGAGCTCAGATTGAAAAGGAGGGAGCTACTCTCAGGCTGCAATGTGAAACAGCTGCCCTGTGTGGGACTGAGTGGCAAGTCCCTTTGTGACTTCAAGAA 85 86 uc003nnw.2,uc003nny.3,uc003nnz.3,uc003nou.4,uc003nov.4,uc003raj.3,uc003ran.1,uc003rtl.5,uc010jrn.2,uc011dmb.4,uc021ytv.1,uc021ytx.2,uc032wqa.1,uc032wqb.1 UCSCgenes 41 IFRD1_intron_comp41141_c0_seq1 82-181 GATAGAGGGTAGGCTTCTTTAGCTCATATGTCTAAGCTTTCTATCTTTGAATTACAGTTGAAGTTTAATGATCTAGTAAGCACCTGTAAAGCATACAGAA 73 75 comp41141_c0_seq1 trinity 42 IKBIP_isoform 926-1025 GCAGGAAATAAAATTGCTCACTGAACGGCTAAAAGATTTGGAAGATAGCACACTAAGAAATATTAGAACAGTAAAAAGACAAGAAGAAGAAGATCTCCTG 82 80 uc001tfx.4 UCSCgeneIso 43 IL1R2 114-213 TGCTTCTGCCACGTGCTGCTGGGTCTCAGTCCTCCACTTCCCGTGTCCTCTGGAAGTTGTCAGGAGCAATGTTGCGCTTGTACGTGTTGGTAATGGGAGT 81 78 uc002tbm.3,uc002tbn.4,uc002tbo.2,uc031rop.1;uc002tbm.3 UCSCgenes UCSCgeneIso 44 KIAA1551 746-845 CTTCAGGAGTTACCCAAAACGTATGGTTGAACTCACCAATGAGGAATCCTGTGCATTCTCATATAGGGGCAACTGTATCTCATCAAACTGATTTTGGAGC 81 79 ENSG00000174718 ensembl 45 KRT23 1736-1835 CGGGAAGAATCAAAGTCGAGCATGAAAGTGTCTGCAACTCCAAAGATCAAGGCCATAACCCAGGAGACCATCAACGGAAGATTAGTTCTTTGTCAAGTGA 83 82 uc002hvm.2,uc002hvn.1,uc010cxf.2,uc010cxg.3,uc010wfl.2;ENSG00000108244 UCSCgenes ensembl 46 LYST 179-278 GGTCATGAGCACCGACAGTAACTCACTGGCACGTGAATTTCTGACCGATGTCAACCGGCTTTGCAATGCAGTGGTCCAGAGGGTGGAGGCCAGGGAGGAA 83 83 comp56924_c0_seq196;comp56924_c0_seq84;comp56924_c0_seq7 Trinity Trinity trinity 47 MAP3K8_isoform 25-124 GTCAGTTTCCCATGGGTCTTGAATGCAAATACAAATATCGTAAACTAAATATTTGTGTTTTCTTTCCTAGACTCTCCAGAAAGAGCAACAGTAATGGAGT 74 75 uc001ivj.2 UCSCgeneIso 48 MBNL3 525-624 CAAATACACCTGTTCTGATTCCTGGAAACCCACCTCTTGCAATGCCAGGAGCTGTTGGCCCAAAACTGATGCGTTCAGATAAACTGGAGGTTTGCCGAGA 78 79 uc004ewv.4;comp56305_c1_seq11 UCSCgeneIso trinity 49 MME 5060-5159 GGATTGTAGGTGCAAGCTGTCCAGAGAAAAGAGTCCTTGTTCCAGCCCTATTCTGCCACTCCTGACAGGGTGACCTTGGGTATTTGCAATATTCCTTTGG 80 79 uc003fab.1,uc003fac.1,uc003fad.1,uc003fae.1,uc010hvr.1,uc031sca.1,uc031scb.1;uc003fab.1;ENSG00000UCSCgenes UCSCgeneIso ensembl  197   Position Target Sequence Tm CP Tm RP Target ID Datasets 196549 50 MT-CYB 656-755 CCTTCCACCCTTACTACACAATCAAAGACGCCCTCGGCTTACTTCTCTTCCTTCTCTCCTTAATGACATTAACACTATTCTCACCAGACCTCCTAGGCGA 81 79 ENSG00000198727 ensembl 51 MT-ND1 478-577 TTTAACCTCTCCACCCTTATCACAACACAAGAACACCTCTGATTACTCCTGCCATCATGACCCTTGGCCATAATATGATTTATCTCCACACTAGCAGAGA 80 82 ENSG00000198888 ensembl 52 NAPA 897-996 GTACAGCGCCAAAGACTACTTCTTCAAGGCGGCCCTCTGCCACTTCTGCATCGACATGCTCAACGCCAAGCTGGCTGTCCAAAAGTATGAGGAGCTGTTC 83 84 uc002pha.2,uc002phc.2,uc002phd.2,uc002phe.3,uc010elg.2,uc032iap.1;uc002pha.2;ENSG00000105402 UCSCgenes UCSCgeneIso ensembl 53 NFKB1 1676-1775 AGGGTATAGCTTCCCACACTATGGATTTCCTACTTATGGTGGGATTACTTTCCATCCTGGAACTACTAAATCTAATGCTGGGATGAAGCATGGAACCATG 79 79 comp56220_c0_seq24 trinity 54 NFKBIA 946-1045 GGATGAGGAGAGCTATGACACAGAGTCAGAGTTCACGGAGTTCACAGAGGACGAGCTGCCCTATGATGACTGTGTGTTTGGAGGCCAGCGTCTGACGTTA 82 83 uc001wtf.4;uc001wtf.4;ENSG00000100906 UCSCgenes UCSCgeneIso ensembl 55 PABPC1 322-421 CCAGCGGCAGTGGATCGACCCCGTTCTGCGGCCGTTGAGTAGTTTTCAATTCCGGTTGATTTTTGTCCCTCTGCGCTTGCTCCCCGCTCCCCTCCCCCCG 81 82 uc003yjs.1,uc003yjt.1,uc003yju.2,uc011lhc.1,uc011lhd.1,uc033bur.1;uc011lhd.1;ENSG00000070756 UCSCgenes UCSCgeneIso ensembl 56 PELI1_isoform 178-277 CACCACAAAGCAGCCCCAACGCCTCTCCCTGCGTCCGCGGCTCCTCAGCGCTCGGCTCCGTGGTCAACTTCCCCTCGCTGGGCTCGGCTGGCGGGCGCGG 86 85 uc002sct.4 UCSCgeneIso 57 PLAGL2 5296-5395 TCTCCACCTTTGGCACTAGAACTCCTGAGACACCACTTCTCATGCTTCTCCCTCCCTACCAGCGGTCAAGGCTTTGGAGCCACTCTTTTGTAACTCCAGA 82 82 ENSG00000126003 ensembl 58 PPP3R1 2386-2485 GCCATCGCTGTTCCTTCAACTGAGTGCTGCACATCATGGGCTCTGTCTGTGAGAGAAAAATCCCGGTGCTTGGTGTCCTTGCATGACATGGAGTTTTGCA 83 81 uc002sei.1;uc002sei.1;ENSG00000221823 UCSCgenes UCSCgeneIso ensembl 59 PSMF1 3071-3170 TTCCTGGTCAAGCTTATGGTGCTTATTTTGATCTGGGCCACTTCCCTCCTTCCAGTCATGAGTAATCATCAAGGAGCAAGTTGGAGTGTTTCAGGTGTAT 79 78 uc002wel.4,uc002wen.4,uc002wep.4,uc010zpo.2,uc010zpp.2;uc002wen.4;ENSG00000125818;comp49646_c0_seq1 UCSCgenes UCSCgeneIso Ensemble trinity 60 PTAR1_isoform 742-841 TTTCTACCTTCAGCATCACTTAAATGGTAGGTTTCCTCACAGCATGACCCAGTTGTCACCTGCAGACAGCCCT 78 86 uc004ahi.3 UCSCgeneIso  198   Position Target Sequence Tm CP Tm RP Target ID Datasets GGGGGGACTTTGAGTGACTTGCACCTT 61 PTPN18 3421-3520 CCTCTGTGTTGCTGGATAATGAGTCATCTATCTCTGGAGGAGAAGAAAGGCAGGTCCTCCACAGCCCTGATAAAATCTCCAAGTCTCCCAGTTTCGGGTC 84 84 uc002trb.3,uc002trc.3,uc002tre.3;uc002trc.3 UCSCgenes UCSCgeneIso 62 QKI 839-938 TAATTTTGTTGGGAGAATCCTTGGACCTAGAGGACTTACAGCCAAACAACTTGAAGCAGAAACCGGATGTAAAATCATGGTCCGAGGCAAAGGCTCAATG 82 83 comp56263_c2_seq15 trinity 63 RAB5B 2636-2735 CAGACTGCCTTCTATCCCAGAACAGCTGAGAAATCTATGAAGCTGAGATTCTGAAGGACCCAGCTTAGGTTCTTCCACTTAGGCCTCAATTCCCTTCCTT 82 80 comp40093_c0_seq1 trinity 64 RALGPS2 1486-1585 GAGGCCAAGCTGAAAGTTCTACTCTTTCTAGTGGAATATCAATAGGTAGCAGCGATGGTTCTGAACTAAGTGAAGAGACCTCATGGCCTGCTTTTGAAAG 81 83 ENSG00000116191 ensembl 65 RGS2 856-955 AACAGCTTCCCTCACTGTGTACAGAACGCAAGAAGGGAATAGGTGGTCTGAACGTGGTGTCTCACTCTGAAAAGCAGGAATGTAAGATGATGAAAGAGAC 82 76 uc001gsl.3;uc001gsl.3;ENSG00000116741;comp54861_c0_seq12 UCSCgenes UCSCgeneIso Ensemble trinity 66 SEMA4D 1121-1220 AAGTGAACCCATCATCTCCCGAAATTCTTCCCACAGTCCTCTGAGGACAGAATATGCAATCCCTTGGCTGAACGAGCCTAGTTTCGTGTTTGCTGACGTG 84 79 uc004aqo.1 UCSCgeneIso 67 SEPT7 3773-3872 TCCTGTTAGCCACTGTCTTTCTGCTAATTAAGTGGGGCTGAACAAGTAAGCACTAATAATACCAGTGAACCACTTGGGCACCTTGTGGGTAGAGTTTTGC 81 82 ENSG00000122545 ensembl 68 SETX 1141-1240 ATCAACAACGCAAGCTACAATAGAGAGATCCGACATATACGGAACAGCTCTGTAAGGACCAAGTTAGAACCGGAGTCCTATCTGGATGATATGGTGACTT 80 80 comp56690_c2_seq36 trinity 69 SF3B1 1-100 GGAAGTTCTTGGGAGCGCCAGTTCCGTCTGTGTGTTCGAGTGGACAAAATGGCGAAGATCGCCAAGACTCACGAAGATATTGAAGCACAGATTCGAGAAA 82 83 uc002uue.3,uc002uuf.3,uc002uug.3,uc010fsk.1;uc002uue.3;ENSG00000115524;comp56763_c0_seq7 UCSCgenes UCSCgeneIso Ensemble trinity 70 SH3BGRL3 391-490 AGATTGTCAACGGGGACCAGTACTGTGGGGACTATGAGCTCTTCGTGGAGGCTGTGGAACAAAACACGCTGCAGGAGTTCCTGAAGCTGGCTTGAGTCAA 84 82 uc001blu.3;ENSG00000142669 UCSCgenes ensembl 71 SLC35E2B 1470-1569 CCGGTGACTTTCAGCGTCGCCAGCACCGTGAAACATGCCTTGTCCATCTGGCTCAGCGTAATCGTTTTCGGCAACAAGATCACCAGCTTGTCGGCCGTTG 84 79 ENSG00000189339 ensembl 72 SMCHD1 4676-4775 AACCACCTACACCAGCTGTTTCAAATGTTCGCTCAGT 84 81 ENSG00000101596 ensembl  199   Position Target Sequence Tm CP Tm RP Target ID Datasets TGCCAGTAGGACCTTGGTCAGAGATCTACATCTTAGTATCACGGATGACTACGACAACCATAC 73 SPARC 911-1010 TTTTCGAGACCTGTGACCTGGACAATGACAAGTACATCGCCCTGGATGAGTGGGCCGGCTGCTTCGGCATCAAGCAGAAGGATATCGACAAGGATCTTGT 83 82 uc003lug.4,uc003lui.4;ENSG00000113140 UCSCgenes ensembl 74 SULT1A1 1394-1493 TGCGAATCAAACCTGACCAAGCGGCTCAAGAATAAAATATGAATTGAGGGCCCGGGACGGTAGGTCATGTCTGTAATCCCAGCAATTTGGAGGCTGAGGT 84 85 uc002dqi.3,uc002dqj.3,uc002dqk.3,uc002dql.3,uc002dqm.3,uc002dqn.3,uc002dqp.3;uc002dql.3;ENSG00000196502 UCSCgenes UCSCgeneIso ensembl 75 TEX261 2741-2840 TCAAAACGGTCAGGTCTACCTTAACATCTCTTGATTTGAGCCACTCCCACTGTCATCAGCTTTCACCTGGATTATCGTGACAGCCTCCTACTGCTTCTCT 81 80 ENSG00000144043 ensembl 76 TGFBI 2031-2130 GTGGTCCATGTCATCACCAATGTTCTGCAGCCTCCAGCCAACAGACCTCAGGAAAGAGGGGATGAACTTGCAGACTCTGCGCTTGAGATCTTCAAACAAG 82 80 uc003lbf.4,uc003lbg.4,uc003lbh.4,uc010jee.3,uc011cyb.2;uc003lbf.4;ENSG00000120708 UCSCgenes UCSCgeneIso ensembl 77 TIA1_isoform 883-982 GACTTATTGCAGAAATAGATGAGAAGCAAATCAAGACTACTATTCAAAAATCAAATACCAAACAGCTATCATATGATGAGGTTGTAAATCAGTCTAGTCC 77 77 uc002sgl.4 UCSCgeneIso 78 TPP1 747-846 AGTTCCTGGAGCAGTATTTCCATGACTCAGACCTGGCTCAGTTCATGCGCCTCTTCGGTGGCAACTTTGCACATCAGGCATCAGTAGCCCGTGTGGTTGG 84 83 uc001mek.1,uc001mel.1,uc010rar.1;ENSG00000166340;comp50683_c0_seq2 UCSCgenes Ensemble trinity 79 UBE2D1 549-648 TACTTTGTGATCCTAATCCAGATGACCCCTTAGTACCAGATATTGCACAAATCTATAAATCAGACAAAGAAAAATACAACAGACATGCAAGAGAATGGAC 78 83 uc001jke.2,uc021prc.1;ENSG00000072401 UCSCgenes ensembl 80 unknown1_comp54405_c1_seq1 5-104 TTCTTATCGTGGTGGTAGTTCCACAATGTATACACACGTCTCAATCTACCAAACTACATGCTTCAAAAATGTACACTTTGTTTCATGCCCATTATACCTC 76 73 comp54405_c1_seq1 trinity 81 unknown2_comp55647_c0_seq2 1-100 AAGGGACTGGGTTTCATCACCAGTGACTGCCATGAGTCTAACTTAGTTATCTGGTTGGATAGGGATGCATTTACTGAACTGGGGACCCAGGAGGAGTTGG 82 85 comp55647_c0_seq2 trinity 82 unknown2a_comp55647_c0_seq2 496-595 GCTGCACATCAAAATCACTGTTTCCTTTGACCAATACTCTAGACACACTCAAGCAGAGATTCTGGATTCTGATTTAATTGACACAGAGTAGGGCCTGGGC 79 84 comp55647_c0_seq2 trinity 83 unknown3_comp56590_c0_seq8 16-115 GCTGCCCCGAGCCCGCGAAGGGAGGGAAGTTCCAGAATCGAGAGAGGGAGGGAGTCAAGGTGGAACCCATAGAGTGAGCCTCCTGAAGACACAGAGCGGT 87 87 comp56590_c0_seq8 trinity  200   Position Target Sequence Tm CP Tm RP Target ID Datasets 84 VPS13A_isoform 9413-9512 CAGGTTGAGGGATGGGACTGGAAATCAAATGTTACAGGCATCAAAAAGTTTGATATGAAAAGTTAATGCATGACTTTGCAAGTGAAAGCCAACAGTAGAT 83 78 uc004akp.4 UCSCgeneIso 85 ZNF185 1124-1223 CTGAAGGCTTGGCTGCAGTAGACATCGGCTCCGAGAGAGGAAGCTCCAGTGCCACTTCAGTCTCTGCTGTCCCTGCTGATAGGAAGAGCAACAGCACAGC 85 84 uc004fgu.3,uc004fgv.3,uc004fgw.4,uc004fgx.3,uc010ntv.2,uc011myg.2,uc011myh.2,uc011myi.2,uc011myj.2,uc011myk.2;ENSG00000147394 UCSCgenes ensembl 86 ZNF281 2606-2705 AGCGTTTGGTTCTCAGTTTAAGTCGGGCAGCAGGGTGCCAATGACCTTTATCACTAACTCTAATGGAGAAGTGGACCATAGAGTAAGGACTTCAGTGTCA 80 84 ENSG00000162702 ensembl 87 ZNF609_isoform 757-856 CCTGTTTCCACACCAGCAGTGCTGCCAATACACCTTTTGGTGCCAGTGGTCAACAATGACATCTCATCTCCTTGTGAGCAGATCATGGTTCGTACCCGAT 84 79 uc002ann.3 UCSCgeneIso Tm CP: melting temperature of the capture probe Tm RP: melting temperature of the reporter probe   201 C.8! Trinty contigs mapping to known genes  Contig ID Sequence length (bp) Query cover (%) Identity (%) Chromosome Gene 1 comp55716_c0_seq31 756 100 100 2 CASP8 100 100 100 2 comp55863_c0_seq6 1,986 100 100 22 CECR1 100 100 100 3 comp57541_c0_seq1 747 100 99 mitochondrion CYTB 100 100 100 4 comp56357_c0_seq8 2,213 99 99 5 FNIP1 100 100 100 5 comp55353_c2_seq11 2,433 99 99 19 FPR2 100 100 100 6 comp56950_c0_seq1 606 94 99 16 HBA2 100 100 100 7 comp54057_c0_seq6 3,899 97 99 3 HCLS1 100 93 100   8 comp56924_c0_seq196 10,434 100 100 1   LYST  100 100 100 comp56924_c0_seq84 9,263 100 100 1 100 100 100 comp56924_c0_seq7 12,791 100 100 1 100 100 100 9 comp56305_c1_seq11 3,853 98 100 X MBNL3 100 91 100 10 comp56220_c0_seq24 3,397 100 100 4 NFKB1 100 93 100 11 comp49646_c0_seq1 3,506 99 99 20 PSMF1 100 100 100 12 comp56263_c2_seq153 3,356 99 99 6 QKI 100 100 100 13 comp40093_c0_seq1 1,735 100 100 12 RAB5B 100 100 100 14 comp54861_c0_seq12 1,368 99 100 1 RGS2 100 100 100 15 comp56690_c2_seq36 5,606 100 99 9 SETX 100 100 100 16 comp56763_c0_seq7 4,335 99 100 2 SF3B1 100 76 100 17 comp50683_c0_seq2 1,317 98 100 11 TPP1 100 100 100     202  C.9! Representative example of probe design for annotated contigs: comp54057_c0_seq6 (HCLS1) RNA-Sequencing: Mapping of the contig sequence to HCLS1 (see below for contig expression)  NanoString: Mapping of 100 bp probe for HCLS1 (see below for probe expression)     121,660 K121,655 K121,650 K121,645 K121,640 K121,635 K121,630 KSequenceFBXO40HCLS1 GOLGB1GenesQuery_6...BLAST Results for: Nucleotide Sequence (3899 letters)Query_67965Cleaned Alignments - BLAST Results for: Nucleotide Sequence (3899 letters)NC_000003.12:121629916..121665188 Homo sapiens chromosome 3, GRCh38.p2 Primary Assembly121,660 K121,655 K121,650 K121,645 K121,640 K121,635 KSequenceHCLS1GenesBLAST Results for: Nucleotide Sequence (100 letters)Query_219039Cleaned Alignments - BLAST Results for: Nucleotide Sequence (100 letters)NC_000003.12:121630544..121661783 Homo sapiens chromosome 3, GRCh38.p2 Primary Assemblyp=0.01Discovery4567DR ERResponselog2 cpm (RNA−sequencing)●p=0.11Recalibration10.7511.0011.2511.5011.75DR ERResponselog2 counts (NanoString)●p=0.64p=0.0006p=0.02p=0.21p=0.0004p=0.004Validation10.511.011.512.0HC (9am)HC (12pm) DR ERResponselog2 Counts (NanoString) 203 C.10! Trinty contigs mapping to intronic regions or uncharacterized loci  Contig ID Sequence length (bp) Query cover (%) Identity (%) chromosome Gene/Locus 1 comp56590_c0_seq8 1,360 57 99 8 TNFRSF10C_intron LOC254896_exon 100 100 100 100 100 100 2 comp17070_c0_seq1 947 100 97 19 FPR1_intron 100 100 100 3 comp55647_c0_seq2 3,130 57 99 3 LOC101927568_intron 100 2 – 100 2 – 100 1 100 2a – 100 2a – 100 1 LOC105378945_intron 4 comp54405_c1_seq1  2,368 4 96 17 SSH2_intron 100 100 100 8 TNFRSF10C_intron 5 comp41141_c0_seq1 1,462 91 99 7 IFRD1_intron 100 100 100      204 C.11! Coverage plots and genome alignment of Trintiy contigs that mapped to intronic or uncharacterized loci For each uncharacterized contig the following are shown: A.! Contig mapping to the human genome B.! Mean ± SD coverage for each base of the contig for ERs and DRs. Blue line represents the 100 bp sequence chosen for the NanoString platform based on visual inspection. C.! 100 bp mapping to the human genome D.! Boxplots displaying: left: contig expression in ERs and DRs in the discovery cohort (36 samples), Middle: nanoString probe expression in ERs and DRs in the same samples as the discovery cohort (29/36 samples), Right: nanoString probe expression in healthy controls (at 9am and 12pm) and ERs and DRs in the independent external cohort (45 samples)    205 comp56590_c0_seq8 (TNFRSF10C_intron, LOC254896_exon)  A.! Mapping of contig sequence to the human genome   B.! Coverage plot   C.! Mapping of the nanoString probe to the human genome    23,125 K23,120 K23,115 K23,110 K23,105 K23,100 K23,095 K23,090 K23,085 K23,080 KSequenceLOC286059LOC254896NR_046173.1TNFRSF10CNM_003841.3 NP_003832.2GenesBLAST Results for: Nucleotide Sequence (1360 letters)Cleaned Alignments - BLAST Results for: Nucleotide Sequence (1360 letters)NC_000008.11:23076244..23130900 Homo sapiens chromosome 8, GRCh38.p2 Primary Assembly●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●0501001500 500 1000Contig sequenceCoverage Phenotype●●ERDRcomp56590_c0_seq8(LOC254896_exon.TNFRSF10C_intron)   length = 1360 bases23,125 K23,120 K23,115 K23,110 K23,105 K23,100 K23,095 K23,090 K23,085 K23,080 K23,075 KSequenceLOC286059LOC254896NR_046173.1TNFRSF10CNM_003841.3 NP_003832.2GenesBLAST Results for: Nucleotide Sequence (100 letters)NC_000008.11:23074777..23132121 Homo sapiens chromosome 8, GRCh38.p2 Primary Assembly 206 D.! Boxplots of expression in healthy controls and asthmatics in discovery, recalibration and validation cohorts     ●●p=0.0003Discovery4.04.55.0DR ERResponselog2 cpm (RNA−sequencing)p=0.24Recalibration10.010.511.0DR ERResponselog2 counts (NanoString)●p=0.16p=0.03p=7.3x10^−5p=0.23p=0.01p=0.0004Validation101112HC (9am) HC (12pm) DR ERResponselog2 Counts (NanoString) 207 comp17070_c0_seq1 (FPR1_intron)  A.! Mapping of contig sequence to the human genome  B.! Coverage plot   C.! Mapping of the nanoString probe to the human genome      D.! Boxplots of expression in healthy controls and asthmatics in discovery, recalibration and validation cohorts 51,752 K51,751 K51,750 K51,749 K51,748 K51,747 K51,746 K51,745 KSequenceFPR1GenesQuery_9725BLAST Results for: Nucleotide Sequence (947 letters)NC_000019.10:51744418..51752675 Homo sapiens chromosome 19, GRCh38.p2 Primary Assembly●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●02040600 250 500 750Contig sequenceCoverage Phenotype●●ERDRcomp17070_c0_seq1(FPR1_intron)   length = 947 bases51,753 K51,752,50051,752 K51,751,50051,751 K51,750,50051,750 K51,749,50051,749 K51,748,50051,748 K51,747,50051,747 K51,746,50051,746 K51,745,500SequenceFPR1GenesBLAST Results for: Nucleotide Sequence (100 letters)NC_000019.10:51745471..51753440 Homo sapiens chromosome 19, GRCh38.p2 Primary Assembly 208     ●●p=0.005Discovery3.54.04.5DR ERResponselog2 cpm (RNA−sequencing)●p=0.23Recalibration8.59.09.510.0DR ERResponselog2 counts (NanoString)●p=0.06p=0.18p=0.04p=0.74p=0.15p=0.02Validation8.59.09.510.010.5HC (9am)HC (12pm) DR ERResponselog2 Counts (NanoString) 209 comp55647_c0_seq2 (LOC101927568_intron, LOC105378945_intron)  A.! Mapping of contig sequence to the human genome   B.! Coverage plot   C.! Mapping of the nanoString probes to the human genome comp55647_c0_seq2 (2-LOC101927568_intron)   comp55647_c0_seq2 (2a-LOC105378945_intron) 134,609 K134,608 K134,607 K134,606 K134,605 K134,604 K134,603 K134,602 K134,601 K134,600 K134,599 K134,598 K134,597 K134,596 K134,595 K134,594 KSequenceLOC101927568KYGenesQuery_84965BLAST Results for: Nucleotide Sequence (3130 letters)NC_000003.12:134593938..134609853 Homo sapiens chromosome 3, GRCh38.p2 Primary Assembly●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●2a 20501000 1000 2000 3000Contig sequenceCoverage Phenotype●●ERDRcomp55647_c0_seq2(2−LOC101927568_intron, 2a−LOC105378945_intron)   length = 3130 bases120,915 K120,910 K120,905 K120,900 K120,895 K120,890 K120,885 K120,880 K120,875 K120,870 K120,865 K120,860 KSequenceLOC105378945XR_947782.1PPIAL4BNM_001143883.2 NP_001137355.1 LINC00623GenesBLAST Results for: Nucleotide Sequence (100 letters)Cleaned Alignments - BLAST Results for: Nucleotide Sequence (100 letters)NC_000001.11:120859003..120922769 Homo sapiens chromosome 1, GRCh38.p2 Primary Assembly 210   D.! Boxplots of expression in healthy controls and asthmatics in discovery, recalibration and validation cohorts comp55647_c0_seq2 2-LOC101927568_intron              2a-LOC105378945_intron    134,608 K134,606 K134,604 K134,602 K134,600 K134,598 K134,596 K134,594 KSequenceLOC101927568KYGenesBLAST Results for: Nucleotide Sequence (100 letters)NC_000003.12:134593169..134609110 Homo sapiens chromosome 3, GRCh38.p2 Primary Assembly●p=0.0001Discovery4.85.25.6DR ERResponselog2 cpm (RNA−sequencing)●p=0.27Recalibration7.68.08.4DR ERResponselog2 counts (NanoString)●●p=0.66p=0.10p=0.16p=0.78p=0.08p=0.13Validation89HC (9am) HC (12pm) DR ERResponselog2 Counts (NanoString)●p=0.0001Discovery4.85.25.6DR ERResponselog2 cpm (RNA−sequencing)●p=0.27Recalibration7.68.08.4DR ERResponselog2 counts (NanoString)●●p=0.66p=0.10p=0.16p=0.78p=0.08p=0.13Validation89HC (9am) HC (12pm) DR ERResponselog2 Counts (NanoString)●p=0.0001Discovery4.85.25.6DR ERResponselog2 cpm (RNA−sequencing)p=0.33Recalibration6.57.07.58.08.5DR ERResponselog2 counts (NanoString)●p=0.70p=0.19p=0.22p=0.33p=0.34p=0.43Validation789HC (9am) HC (12pm) DR ERResponselog2 Counts (NanoString) 211 comp54405_c1_seq1 (TNFRSF10C_intron)  A.! Mapping of contig sequence to the human genome   B.! Coverage plot   C.! Mapping of the nanoString probe to the human genome   D.! Boxplots of expression in healthy controls and asthmatics in discovery, recalibration and validation cohorts 30 M29,950 K29,900 K29,850 K29,800 K29,750 K29,700 K29,650 K29,600 KSequenceABHD15NM_198147.2GIT1ANKRD13BCORO6SSH2RPL21P123exonLOC100421100RPL9P30exonEFCAB5RNY4P13GenesBLAST Results for: Nucleotide Sequence (2368 letters)NC_000017.11:29557260..30067590 Homo sapiens chromosome 17, GRCh38.p2 Primary Assembly●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●0501000 500 1000 1500 2000Contig sequenceCoverage Phenotype●●ERDRcomp54405_c1_seq1(TNFRSF10C_intron)   length = 2368 bases23,116 K23,114 K23,112 K23,110 K23,108 K23,106 K23,104 K23,102 KSequenceLOC254896TNFRSF10CNM_003841.3 NP_003832.2GenesBLAST Results for: Nucleotide Sequence (100 letters)Cleaned Alignments - BLAST Results for: Nucleotide Sequence (100 letters)NC_000008.11:23101257..23116876 Homo sapiens chromosome 8, GRCh38.p2 Primary Assembly 212     ●●●●p=0.0002Discovery5.66.06.46.8DR ERResponselog2 cpm (RNA−sequencing)p=0.008Recalibration9.09.510.010.5DR ERResponselog2 counts (NanoString)●p=0.04p=0.04p=0.0001p=0.13p=0.008p=0.0001Validation9.09.510.010.5HC (9am)HC (12pm) DR ERResponselog2 Counts (NanoString) 213 comp41141_c0_seq1 (IFRD1_intron)  A.! Mapping of contig sequence to the human genome   B.! Coverage plot   C.! Mapping of the nanoString probe to the human genome     D.! Boxplots of expression in healthy controls and asthmatics in discovery, recalibration and validation cohorts 112,500 K112,490 K112,480 K112,470 K112,460 K112,450 K112,440 K112,430 K112,420 KSequenceLOC105375457IFRD1LSMEM1GenesBLAST Results for: Nucleotide Sequence (1462 letters)NC_000007.14:112414195..112508147 Homo sapiens chromosome 7, GRCh38.p2 Primary Assembly●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●