Open Collections

DNA methylation in human development : methodologies and analytics for genome-wide studies

University of British Columbia

High-throughput methods have resulted in a large volume of studies measuring genome-wide DNA methylation (DNAm) in association with human health and disease. Understanding of DNAm patterns may be translated, for example, into predicting children at risk for illness or identifying etiological subtypes within a heterogeneous disease. Addressing biological and technical factors affecting measurement of genome-wide DNAm is essential to reduce false discovery in such studies. This dissertation develops principles for analyzing genome-wide DNAm, with the aim of improving collection and analysis of human developmental data. To this end, I present four studies employing several techniques to measure genome-wide DNAm: DNAm of L1 and Alu repetitive elements in addition to Illumina 27k and 450k DNAm microarrays. In the first of these studies, I found that tissue type, gestational age, technical platform and CpG density contribute to variable measurement of genome-wide DNAm. Subsequent studies primarily used the 450k array to measure genome-wide DNAm, a technology targeting 485,577 sites in the genome. A detailed annotation of the 450k array was created and tested, to enhance this platform’s utility. Array probes targeting sites containing SNPs (4.3%) and non-specific probes (8.6%) were identified, and I examined how these compromised probes may result in spurious discoveries. A pilot study in placental tissue identified batch effects in 450k data. A computational tool was applied to reduce the batch signal, but I demonstrated that when applied to a problematic study design, false biological signal may be introduced. The workflow for processing and analyzing genome-wide DNAm data was finally applied to profile five tissues ascertained from second trimester neural tube defect (NTD)-affect pregnancies. Despite research, medical interventions and public health changes, NTDs remain the second most common congenital abnormality in many parts of the world, and the etiology of these cases is unknown. Using the 450k array, I found 3,342 differentially methylated sites in the kidneys of spina bifida cases compared to gestational-age matched controls, but little alteration in genome-wide DNAm in other NTD tissues. This dissertation contributes methodologies and analytical tools that will help manage bias, improve reproducibility and reduce false discoveries in studies of genome-wide DNAm.

Text

2016-04-26

Attribution-NonCommercial-NoDerivatives 4.0 International

http://hdl.handle.net/2429/57845

Reproductive and Developmental Sciences

University of British Columbia

UBCV

http://creativecommons.org/licenses/by-nc-nd/4.0/