UBC Faculty Research and Publications

Transposon fingerprinting using low coverage whole genome shotgun sequencing in Cacao (Theobroma cacao… Sveinsson, Saemundur; Gill, Navdeep; Kane, Nolan C; Cronk, Quentin Jul 24, 2013

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

Item Metadata


52383-12864_2013_Article_5202.pdf [ 469.59kB ]
JSON: 52383-1.0223604.json
JSON-LD: 52383-1.0223604-ld.json
RDF/XML (Pretty): 52383-1.0223604-rdf.xml
RDF/JSON: 52383-1.0223604-rdf.json
Turtle: 52383-1.0223604-turtle.txt
N-Triples: 52383-1.0223604-rdf-ntriples.txt
Original Record: 52383-1.0223604-source.json
Full Text

Full Text

RESEARCH ARTICLE Open AccessTransposon fingerprinting using low coveragewhole genome shotgun sequencing in Cacao(Theobroma cacao L.) and related speciesSaemundur Sveinsson1*, Navdeep Gill1, Nolan C Kane2 and Quentin Cronk1AbstractBackground: Transposable elements (TEs) and other repetitive elements are a large and dynamically evolving partof eukaryotic genomes, especially in plants where they can account for a significant proportion of genome size.Their dynamic nature gives them the potential for use in identifying and characterizing crop germplasm. However,their repetitive nature makes them challenging to study using conventional methods of molecular biology. Nextgeneration sequencing and new computational tools have greatly facilitated the investigation of TE variation withinspecies and among closely related species.Results: (i) We generated low-coverage Illumina whole genome shotgun sequencing reads for multiple individualsof cacao (Theobroma cacao) and related species. These reads were analysed using both an alignment/mappingapproach and a de novo (graph based clustering) approach. (ii) A standard set of ultra-conserved orthologoussequences (UCOS) standardized TE data between samples and provided phylogenetic information on therelatedness of samples. (iii) The mapping approach proved highly effective within the reference species butunderestimated TE abundance in interspecific comparisons relative to the de novo methods. (iv) Individual T. cacaoaccessions have unique patterns of TE abundance indicating that the TE composition of the genome is evolvingactively within this species. (v) LTR/Gypsy elements are the most abundant, comprising c.10% of the genome. (vi)Within T. cacao the retroelement families show an order of magnitude greater sequence variability than the DNAtransposon families. (vii) Theobroma grandiflorum has a similar TE composition to T. cacao, but the related genusHerrania is rather different, with LTRs making up a lower proportion of the genome, perhaps because of a massivepresence (c. 20%) of distinctive low complexity satellite-like repeats in this genome.Conclusions: (i) Short read alignment/mapping to reference TE contigs provides a simple and effective method ofinvestigating intraspecific differences in TE composition. It is not appropriate for comparing repetitive elementsacross the species boundaries, for which de novo methods are more appropriate. (ii) Individual T. cacao accessionshave unique spectra of TE composition indicating active evolution of TE abundance within this species. TE patternscould potentially be used as a “fingerprint” to identify and characterize cacao accessions.Keywords: Theobroma cacao, Transposable elements, Next generation sequencing, Graph based clustering,Retrotransposon* Correspondence: saemundur.sveinsson@gmail.com1Department of Botany and Biodiversity Research Centre, University of BritishColumbia, 6270 University Boulevard, Vancouver, BC V6T 1Z4, CanadaFull list of author information is available at the end of the article© 2013 Sveinsson et al.; licensee BioMed Central Ltd. This is an Open Access article distributed under the terms of the CreativeCommons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, andreproduction in any medium, provided the original work is properly cited.Sveinsson et al. BMC Genomics 2013, 14:502http://www.biomedcentral.com/1471-2164/14/502BackgroundTransposable elements (TEs) are a large and dynamicallyevolving part of plant genomes [1,2]. They occupy between15% - 84% of plant genomes [3] and TE expansion isknown to cause a significant increase in genome sizein many cases [4]. Transposable elements are a majorforce in plant evolution, not only by causing genomeexpansions but also by altering gene function eitherthrough disruption [5] or acting as a raw material fornew genes and novel functions [6,7].Transposable elements are usually classified into twomajor classes based on their transposition mechanisms.Class I retrotransposons move about in a ‘copy-and-paste’fashion, through a RNA intermediate, which is encodedback into DNA by an endogenous Reverse Transcriptase(RT) enzyme [8]. The two largest super-families ofretrotransposons in plants, the LTR/Copia and LTR/Gypsy,have several other open reading frames, which play a rolein the transposition, located between two regions of longterminal repeats (LTR) [1]. Class II DNA elements moveabout in genomes through a DNA intermediate. Themost extensively studied group of class II elementstranspose by a ‘cut-and-paste’ mechanism and are classifiedinto several super-families based on sequence similarity[9]. Cut-and-paste DNA transposons are characterized bya transposase gene and a pair of flanking terminal invertedrepeats (TIRS) [7].Transposable elements are known to vary extensivelyin copy-number and nucleotide sequence among closelyrelated species [4,10] and even within the same species[11]. Plant LTR retrotransposons are well known to haveintraspecific variation in copy-number [12,13]. This, incombination with the easily amplifiable LTR domain, hasbeen used in the development of molecular markers forseveral crop species [14-16]. In addition to the extensivepresence/absence variability of the LTR elements, sequenceheterogeneity is also known to be quite extensive [17]. Thereverse transcriptase domain is the most extensively studiedretrotransposon gene and it is known to show levels ofheterogeneity from about 5% to 75% at the amino acidlevel [17]. Heterogeneity and sequence evolution of classII DNA transposons is relatively less studied, but a recentstudy shows that they can be quite heterogeneous [10].Cacao (Theobroma cacao L.) is an economically import-ant tree in the mallow family (Malvaceae) [18]. It is widelygrown in tropical regions as the source of cocoa beans forthe manufacture of chocolate [18]. Cacao has long beenknown to be genetically diverse [19] and traditionally threemajor lineages of Cacao varieties have been recognized:Trinitario, Criollo, and Forastero [20]. Recent work basedon a variety of markers, including microsatellites and wholechloroplast genome sequences of several cacao varieties,has confirmed that the Criollo and Forastero groups aretwo distinct genetic lineages while the Trinitario group is ofhybrid origin [21,22]. Cacao has a relatively small genome,estimated to be around 430 Mb and it has a publishedgenome assembly of about 75% of its estimated genomesize [23]. This small genome size can be partly explainedby the relatively low abundance of transposable elements,compared to other angiosperms. TEs comprise only ap-proximately a quarter of the cacao genome [23].In this study we use low-coverage Illumina whole gen-ome shotgun sequencing to investigate the evolutionarydynamics and comparative analysis of 3,500 TE families innine T. cacao varieties and two related species,Theobromagrandiflorum and Herrania balaensis.ResultsSequence coverage estimates and phylogenetic analysisusing the UCOS contigsThe Illumina sequencing yielded 1.7 – 5.9 Gbp of highquality sequence per sample (Table 1). Average coverageof the ultra-conserved orthologous sequences (UCOS),estimated with BWA mapping, varied between 1.8 and9.4X per sample (see Table 1). This value represents arelative measure of per single copy locus sequencingdepth for our libraries. However it is important to notethat this method may slightly underestimate the sequencingcoverage of H. balaensis and T. grandiflorum due to se-quence divergence in the UCOS among the three spe-cies. Furthermore the results using UCOS are consistentwith results using flow cytometry genome size estimates(see Table 1 in [23]) The UCOS data was used tostandardize the TE data between samples to providerelative TE abundance data. It was also used to estimatethe relatedness of the accessions in order to provide anevolutionary framework for TE variation.The UCOS contigs were informative for the phylogeneticanalysis of H. balaensis, T. grandiflorum and nine of theT. cacao varieties (Figure 1). Scavina-6 was excluded fromthe phylogenetic analysis due to low sequencing depth. Thematrix used to construct the phylogeny consisted of 97UCOS contigs with combined length of 20,438 nucleotides.Individual UCOS alignments varied in length, with theshortest alignment being 54 nucleotides and the longest1,473 nucleotides. Herrania balaensis was set as theoutgroup in the analysis which resulted in two mainTheobroma clades (Figure 1). The first clade consists ofonly T. grandiflorum, with a 100% bootstrap support, andthe second consisting of all T. cacao individuals. This is asexpected given that T. grandiflorum and T. cacao are bio-logically distinct species. Within T. cacao there are twowell-supported clades with T. cacao cv. Stahel andAmelonado grouping together and another well-supportedgrouping of T. cacao cv. Pentagonum, ICS06, ICS39,Criollo-22 and B97 (Figure 1). B97 is the variety used in thewhole genome sequencing project of T. cacao [23]. EET-64and ICS01 are unresolved on a polytomy.Sveinsson et al. BMC Genomics 2013, 14:502 Page 2 of 12http://www.biomedcentral.com/1471-2164/14/502Variation in TE abundance using short read mappingThe coverage of the single copy UCOS genes was usedas a baseline for standardization of TE coverage. Usingthis method, copy-numbers of TE superfamilies relativeto the UCOS coverage in the three species were calculated(Figure 2). The intraspecific variation of copy-number inTheobroma cacao is represented by the error bars. TheLTR super-families are the most numerous elements in thegenomes of H. balaensis, T. grandiflorum and T. cacao, aspreviously shown. The difference in relative copy-numberof class I LTR retroelements between the three species isquite striking. Using this method and the estimated genomesizes for the species [23], it can be calculated that LTR/Gypsy and LTR/Copia elements make up 9% and 7% of thegenome respectively in T. cacao. In contrast these make upjust 2% and 2% in T. grandiflorum and 0.6% and 0.5% inHerrania (Table 2). De novo approaches show the lowvalues in the latter species to be artefacts (see next section).Table 1 Sequence summary statisticsName Chloroplast haplotype1 Read length (bp) No. reads No. reads after trimming UCOS coverageEET-64 (T. cacao) Criollo 60 6.6E + 07 6.5E + 07 5.4Criollo-22 (T. cacao) Criollo 60 4.6E + 07 4.2E + 07 4.4Stahel (T. cacao) Criollo 60 6.0E + 07 5.7e + 07 5.3Pentagonum (T. cacao) Criollo 80 5.8E + 07 5.5E + 07 5.2ICS39 (T. cacao) Criollo 80 6.4E + 07 6.0E + 07 5.9Amelonado (T. cacao) Forastero 60 7.0E + 07 6.8E + 07 5.9ICS06 (T. cacao) Forastero 80 6.4E + 07 6.1E + 07 6.3ICS01 (T. cacao) Forastero 60 5.0E + 07 4.9E + 07 4.4Scavina-6 (T. cacao) Forastero 60 3.8E + 07 2.9E + 07 1.8T. grandiflorum (Cupuaçu) na 60 7.2E + 07 6.8E + 07 5.8H. balaensis na 80 7.7E + 07 7.4E + 07 9.4Illumina sequence summary statistics and observed average coverage of the UCOS contigs for Theobroma cacao, T. grandiflorum and Herrania balaensis based onBurrows-Wheeler Aligner (BWA) alignments.H. balaensisT. grandiflorum*****0.001 substitutions/sitePentagonumStahelICS01AmelonadoEET64ICS06B97ICS39Criollo-22Figure 1 Phylogeny of Herrania balaensis, Theobroma grandiflorum and nine of the T. cacao varieties. The phylogenetic tree wasconstructed using partial sequence data of 97 ultra conserved orthologus sequences (UCOS). Theobroma cacao cv. Scavina-6 was excluded fromthe phylogenetic analysis due to low sequencing coverage. Nodes marked with asterisk have high bootstrap support (>90%).Sveinsson et al. BMC Genomics 2013, 14:502 Page 3 of 12http://www.biomedcentral.com/1471-2164/14/502The apparently lower numbers in species distant from thereference are therefore likely due to mapping incompatibil-ity, and this method is therefore not suitable for interspe-cies studies. The same pattern is also observed even whenreads are mapped to the conserved regions of the LTRretrotransposons using less stringent settings in the shortread aligner (Additional file 1). Mapping incompatibility ismost likely attributable to retroelement divergence betweendistant species. This divergence is an important part of gen-ome differentiation between species and potentially hasimplications for speciation and species divergence. On theother hand de novo methods are efficient in identifying anyrepetitive sequence that is either specific to a given speciesor has mutated beyond recognition that could render itunidentifiable using the mapping approach and thereforelikely to be much more accurate for calculating copy num-bers in species outside the reference species (see below).Variation of TE copy number using de novo approachesA major potential problem with studying interspecificvariation using a mapping approach is that the reads fromHerrania and T. grandiflorum are heterologously mappedto the T. cacao genome. Sequence variation betweenspecies affecting mapping quality could potentially havea considerable effect on the apparent frequency of TEs.We therefore employed a de novo approach in addition,using graph-based clustering. The graph based clusteringanalysis (Figure 3) of the short reads shows considerabledifferences in the representation of the major families ofrepetitive elements in the genomes of the three speciesstudied here. Figure 3 shows the four largest clustersgenerated by RepeatExplorer, their identity and genomepercentage. The most striking difference is between thetwo genera, where H. balaensis has two extremely largelow complexity clusters with combined genome percentageof 19.4%. The two Theobroma species are more similar,with their largest clusters containing about 1-2% of thereads used in the graph based clustering. However, cluster2 in T. cacao is largely composed of LTR/Gypsy elements,whereas the top four clusters in T. grandiflorum are allidentified as low complexity elements. Using graphbased clustering it can be calculated that LTR/Gypsyand LTR/Copia elements make up 10% and 5% of theFigure 2 Relative copy-number of transposable elements using reference based mapping. Relative copy-numbers of the TE super-familiesin the three species represented with bar plots. Relative copy-number was calculated by dividing the total coverage of each super-family, withina sample, by the sample’s mean UCOS coverage. The much lower recovery of transposable elements in the other species is apparently due tomapping failure as the graph based clustering indicates that TE copy numbers are comparable in all species. Error bars representstandard deviation and correspond to intraspecific variation.Table 2 LTR retrotransposon frequencies in the threespecies estimated with two different methodsReference based mappingT. cacao T. grandiflorum H. balanensisLTR/Gypsy 9% 2% 0.6%LTR/Copia 7% 2% 0.5%Graph based clusteringT. cacao T. grandiflorum H. balanensisLTR/Gypsy 10% 10% 4%LTR/Copia 5% 9% 7%Comparison of estimated LTR retrotransposon frequencies as percentages ofthe genome, calculated with reference based mapping (upper half) and graphbased clustering (lower half). Within T. cacao there is little discrepancybetween the methods. Heterologous mapping between species producesdifferent results suggesting that graph-based clustering may be moreappropriate for inter-species comparisons (see Discussion).Sveinsson et al. BMC Genomics 2013, 14:502 Page 4 of 12http://www.biomedcentral.com/1471-2164/14/502genome respectively in T. cacao (comparing well withfigures derived from a mapping approach, Table 2),whereas in Herrania they are 4% and 7% and in Theobromagrandiflorum 10% and 9%. This contrasts with the differentfigures arrived at for the other species using the mappingapproach (see above). We conclude from comparing the re-sults of mapping with the de novo approach that mappingquality in interspecific comparison has an important effecton the results. For a complete list of graph layouts of allclusters, see Additional file 2.Intraspecific variation of TE abundance in T. cacao usingshort read mapping and PCAThe data on intraspecific relative abundance of TEsfrom short read mapping of the eight T. cacao acces-sions was analysed using principal coordinates analysis(PCA) (Figure 4). The first two axes include 81% of thevariance. Axis 1 (46%) separates most strongly the Staheland Amelonado varieties from the Criollo22 variety. TEswith the highest loadings on this axis are DNA transposonssuch as DNA/hAT and LTR retrotransposon such as Copia.T. cacao var. CriolloT. grandiflorumCluster 2Cluster 3Cluster 4H. balaensisLow complexity 10.50% (95,253)8.62% (78,167)Low complexityLTR/GypsyLTR/Gypsy1.40% (12,662)1.34% (12,101)Low complexity1.00% (79,453)0.98% (78,285 )Low complexityLow complexityLow complexity0.93% (74,009) 0.87% (69,073)Low complexity2.21% (45,232) LTR/Gypsy1.84% (37,622)Low complexity1.84% (37,498)Low complexity1.68% (34,288)Cluster 1Figure 3 Graph based clustering analysis of repetitive elements in the three species. Graph layouts of the four largest clusters of repetitiveelements detected in the graph based clustering analysis. Herrania balaensis is shown on the left, T. grandiflorum in the middle and T. cacao cv.Criollo on the right. Clusters are ordered by size, with largest at the top and fourth largest at the bottom. Below each graph layout is the class ofthe repetitive element, the genome percentage of each cluster and number of paired reads belonging to it in parentheses. Coloured regions inthe some graphs represent conserved domains identified by RepeatExplorer. A total of 11,243,224 reads were used in the graph based clustering.Sveinsson et al. BMC Genomics 2013, 14:502 Page 5 of 12http://www.biomedcentral.com/1471-2164/14/502Axis 2 (35%) separates most strongly Scavina-6 fromCriollo22. TEs that load most strongly onto this axis areDNA transposons such as Mutator and Harbinger andGypsy retrotransposons. This analysis clearly separatesmost of the T. cacao samples and shows several clustersof cacao accessions with the first two axes, some ofwhich accord well with phylogenetic relatedness. Due toconcerns that these results might be due to a samplingartefact, particularly given that Scavina-6 has the lowestcoverage, the PCA analysis was repeated individually on49 sub-sampled datasets that were equal in size. However,the same general pattern described above was alwaysobserved (Additional file 3).Sequence conservation of transposable elements in T. cacaoMapping of the short reads of the nine T. cacao librariesto the TE reference contigs revealed considerable levelsof within species nucleotide variability, as calculated bynumber of nucleotide variants detected, divided by thelength of the reference contigs. The nucleotide variabilityin the class II DNA transposons was around 10 times lessthan in the class I retroelements (Figure 5). These resultsillustrate different modes of sequence evolution in classI retroelements and class II DNA elements in the T. cacaostudied here.Comparing the nucleotide variability in two classes ofTEs is informative with regard to how these elementsevolve on the whole but it sheds no light on what partsof individual elements are causing these differences.LTRDigest was able to identify characteristic features ofLTR elements in 355 of the 650 class I families identifiedin [23]. In those 355 families, 90% of the nucleotide vari-ability lay outside of protein coding genes and the longterminal repeats (LTRs), while 5% was in situated withinthe LTRs and 5% in genes (Figure 6). Reverse transcriptase(RT) was the largest contributor, containing about 3% oftotal nucleotide variability followed by integrase withabout 1.5% and the three remaining genes contributedall less than 1% (see Figure 6). However these values areonly informative of total variation not rates of variation,because they differ both in length and representation.LTRDigest does not identify all features in all the ele-ments it interrogated. A better representation of thevariability of the LTR genes and the long terminal repeatis to divide the number of nucleotide variants by thelength of each feature, which yields a comparable esti-mate to the previously calculated nucleotide diversity.Those calculations show that the genes and the longterminal repeat all share a similar value, ranging from about0.002 to 0.09, which are similar values to the averageFigure 4 PCA of the transposable element composition in the Theobroma cacao genotypes. A biplot from a principal component analysis(PCA) using the standardized abundance of each TE super-family as explanatory variables. Percentage of the explained variance is shown inparentheses in the legend of the x- and y-axis.Sveinsson et al. BMC Genomics 2013, 14:502 Page 6 of 12http://www.biomedcentral.com/1471-2164/14/502nucleotide diversity of the class I LTR retroelementsshown in Figure 5.DiscussionThe study of transposable elements (TEs) has been revolu-tionized by the increased availability and lowered costs ofnext generation sequencing (NGS) technologies [24]. NGSmethods have not only been applied in TE studies of plantswith high quality whole genomic sequences available suchas Zea luxurians [25] and rice (Oryza sativa) [26] but alsoin organisms with limited genomic resources available suchas barley (Hordum vulgare), pea (Pisum sativum) and ba-nana (Musa acuminata) [27-29]. These studies demonstratea strong correlation between copy-number estimation ofTEs by traditional molecular methods and methods thatcount short reads from NGS experiments [25,27]. It wastherefore not surprising that the copy-number estimationof TEs in this study fitted very well with previously pub-lished estimates in T. cacao, both in regard to the overallTE abundance in the genome, around 23%, and in thecopy-number of the most abundant class I retroelement[23]. Our study therefore confirms the utility and reliabilityof studying genomic repeats using short reads directly.Different levels of nucleotide conservation in class I andclass II TEsThe two major classes of TE, class I retroelements and classII DNA transposons, have been recognized for a long timeFigure 5 Nucleotide variability of transposable elements in Theobroma cacao. Box plot showing the nucleotide diversity across thesuper-families in T. cacao. This shows that DNA transposons have less variation at the superfamily level (see Discussion). Analyses were performedon standardized data sets (Methods) and values are presented transformed to a log10 scale.ABFigure 6 Nucleotide diversity of LTR/Copia and LTR/Gypsy elements in Theobroma cacao. (A) Schematic diagram of the structure of thetwo most common LTR retrotransposons super-families in the T. cacao genome. (B) Partitioning of nucleotide variation is shown as percentagevalues next to each of the retrotransposon components. The white arrows with black background represents the long terminal repeat (LTR), blackline regions in between open reading frames (ORFs) and LTRs and grey boxes represent the following open reading frames: Reverse transcriptase(RT), integrase (IT), capsid protein (GAG), aspartic proteinase (AP) and Rnase H (RH).Sveinsson et al. BMC Genomics 2013, 14:502 Page 7 of 12http://www.biomedcentral.com/1471-2164/14/502as two fundamentally different groups of mobile elementsprobably present in all eukaryotic genomes [30]. The resultspresented in this study illustrate a considerable differencein the apparent conservation of the TEs in the genomeof T. cacao, where the class I retrotransposons showsignificantly higher levels of heterogeneity, representedby an order of magnitude higher level of nucleotide diver-sity (Figure 5). This may be simply because DNA transpo-sons are more narrowly defined at the superfamily level.However, one possible biological explanation of the highlevels of heterogeneity in class I retroelements results fromtheir transposition mechanism, as described in detail in [7].Class I retrotransposons move about as a RNA intermedi-ate, which is encoded into DNA before re-entry into thehost genome by their endogenous reverse transcriptaseenzyme, which is known to be low-fidelity, causing ahigh mutation rate [7,31].Inter- and intraspecific differences in TE abundance inH. balaensis, T. grandiflorum and T. cacaoTransposable elements are known to cause large inter- andintraspecific differences in the size and composition ofplant genomes, demonstrated in barley (Hordeum vulgare)[11,32] and rice (Oryza sativa) [13]. However, our studyonly found relatively subtle intraspecific differences of theoverall TE abundance in T. cacao. Nevertheless this slightintraspecific variation in TE copy number does potentiallycontribute to the variable genome sizes of different Cacaoaccessions reported in the supplementary material inthe T. cacao genome paper and other sources [23,33,34].Furthermore using a PCA approach to differentiate ac-cessions based on TE abundance, wide separations dooccur (Figure 4). The ability to separate cacao accessionsaccording to TE composition is despite the fact that theyare all closely related, some being of recent hybrid origin[23]. As massive parallel sequencing (MPS) costs fall, thereis interest in using MPS to identify accessions, and such usehas been called “ultra-barcoding” [23]. This paper showsthat data generated for ultra-barcoding could also beused for “transposon composition fingerprinting” of cacaoaccessions (i.e. identification based on a unique spectrumof transposon composition).Mapping vs. de novo approaches to studying TEs fromshort readsOur results (Table 2) suggest that the mapping approach,while reliable within the reference species (T. cacao), isunreliable in interspecific comparisons, at least for someTE families. The mapping approach reports considerabledifferences in the composition of repetitive elements in thethree species studied (Figure 2). Apparently the genomes ofT. grandiflorum and H. balaensis are significantly deficientin many LTR retrotransposons families that are very abun-dant in T. cacao (Figure 3). However this difference may beat least partly caused by low interspecific mapping qualityof the short reads, since our reference contigs originatefrom the genome of T. cacao. The LTR retrotransposonfamilies in particular have high nucleotide diversity(Figure 5), which is likely to cause problems in the mappingof the short reads.The evidence for the failure of the mapping approachin interspecific comparisons comes from the de novoapproach of graph based clustering using RepeatExplorer.This demonstrates that in both T. grandiflorum andH. balaensis the LTR TE families are more abundantthan the mapping approach suggested (Table 2, Figure 3and Additional file 2). More importantly the graph basedclustering showed that the composition of H. balaensis andT. grandiflorum is quite different from T. cacao. Thereforewe conclude that mapping based approaches are well suitedto look at TE evolution in an intraspecific manner whereasde novo methods, such as graph based clustering, are muchmore useful in the exploration of differences in repetitiveelements across species boundaries.ConclusionsThe present study demonstrates considerable differencesin transposable element composition among and evenwithin species, highlighting their dynamic role in plantgenome evolution. Variation of transposable elements inplants is important especially given the great abundanceof transposable elements in plant genomes and their po-tential impact on the genespace. We used two differentmethods of looking at transposable element variationfrom Illumina short read data: reference-based mappingand graph-based clustering. Both are effective at capturingvariation, although each is appropriate at different levelsof taxonomic comparison. Reference based mappingworks well within a species while graph-based clusteringis preferred for between species comparisons.MethodsPlant material and Illumina sequencingTotal genomic DNA was extracted from leaf tissue from11 individuals belonging to three species in the Malvaceae:one Herrania balaensis, one Theobroma grandiflorumand nine T. cacao. Each T. cacao individual representeda different cultivated variety (see Table 1). DNA extractionwas performed using the DNeasy Plant Mini Kit (Qiagen,Valencia, California, USA) according to the manufacturer’sprotocol. Sequencing libraries were constructed usingstandard protocols and chemistry for the Illumina platform.Each library was sequenced on a single lane and generatedeither 60- or 80-bp paired-end sequences (see Table 1)on the Illumina GAII platform by Cofactor Genomicsof St. Louis, MO (http://www.cofactorgenomics.com/).The raw reads are available on NCBI’s Short ReadArchive [SRA048198].Sveinsson et al. BMC Genomics 2013, 14:502 Page 8 of 12http://www.biomedcentral.com/1471-2164/14/502Mapping of reads, coverage estimates and SNP callingThe reference sequences of the transposable element (TE)families used in this study were extracted and characterizedby the authors of the publication describing the T. cacaogenome [23], who graciously made their data available forthis study (Additional file 4). Briefly they identified classI retro-transposons using LTR_finder [35], LTRharvest [36]and in-house software that looked for signatures of classI retroelements, such as the long terminal repeat (LTR) andreverse transcriptase (RT). Class II elements were discov-ered using a blastX search of the transposase gene againstthe Repbase database proteins [37]. In all they identified650 class I - and 2860 class II families. For more details seethe supplementary methods in [23].In order to estimate copy-number and sequence evolu-tion of the TE families using our sequenced libraries ofthree species and nine T. cacao varieties, we mappedreads from each sequenced library to the TE referencecontigs. Firstly, the reads were trimmed for quality, withbases below quality of 20 trimmed from the ends of eachread. Quality trimmed reads were treated as single-endsequences and mapped to the TE reference contigs usingBWA v0.6.1 [38] with the program’s default settings. Therationale behind treating the paired-end sequences as single-end was that TE copy-number estimation from coverage ofthe latter was believed to be more accurate, as paired-endinformation often links the repeat to different single-copyportions of the genome, preventing pairs from mapping nearthe boundaries of the repeated segment. Coverage estimatesfor each nucleotide position in the reference contigs wereextracted from the sorted BAM file output of BWA usingthe genomeCoverageBed tool in the bedTools packagev2.15.0 (genomeCoverageBed flags: -d -ibam) [39]. Relativecopy-number of each TE family was estimated by countingthe number of reads covering each position of the referencecontig and dividing by the length of the contig. Proportionalabundance was calculated for each species, by dividing theabundance of each TE super-family by the abundance of allTEs. Information on nucleotide variants detected in thereads, compared to the TE reference contigs, was extractedusing samtools v0.1.7a [40]. Nucleotide diversity wasestimated for each TE reference contig by counting thenumber of variable sites, with read-depth higher than 6and base qualities higher than 20 (column 6 from samtoolspileup –vcf output), and dividing by the length of thecontig. To control for the effect of different read depths be-tween different libraries, subsampling was used to ensureequality of total reads. Due to the repetitive nature of TEs,a variable site could represent a single nucleotide polymor-phisms in a homologous copy, i.e. a heterozygote, or couldstem from sequence divergence between different copies ofa transposable element.To account for differences in sequencing depth andread length of different libraries, reduced equalized datasets were used for some of the analyses presented here(Figures 4 and 5). The reduced data sets were generatedby trimming the read length of all libraries to 60-bpand randomly extracting reads from all but the smallestsequenced library (Scavina-6). The purpose of this step wasto make sure that variable read lengths and sequencingdepths were not the cause of observed differences in TEcoverage and nucleotide diversity. However any observeddifferences in UCOS coverage could be due to differencesin genome sizes among the three species and the T. cacaovarieties. Furthermore 49 sampling replicates were gener-ated in order to test the effect of data sub-sampling on TEcoverage estimates.The class I LTR retrotransposons reference contigs wereannotated using LTRHarvest [36] and LTRDigest [41].These programs use similarity searches of conserved re-gions of LTR elements, such as the long terminal repeatand protein coding genes, to estimate the coordinates ofthe various features of the elements. That information wasthen used to estimate the variability of each of the LTRelement feature, by combining the feature file output ofLTRDigest with the nucleotide variant output of samtools.In order to test whether we could better account forsequence divergence in the class I TE among the threespecies, we tried mapping the reads exclusively to conservedregions of the LTR retrotransposons and with relaxed BWAalignment stringency. Protein coding regions and the LTRwere extracted from the reference contigs, based on theannotations from LTRHarvest [36] and LTRDigest [41] andthe BWA alignment step was preformed with more relaxedsettings (bwa flags: -l 1024 -i 0 -o 3). These settings allowedfor more gaps and BWA used longer seed length for itsshort read alignments. These relaxed settings and theconserved regions of the LTR elements were only usedto generate Additional file 1.Identification of, and mapping to UCOS contigsA set of 357 Ultra Conserved Orthologous sequences(UCOS, http://compgenomics.ucdavis.edu/compositae_reference.php) was used to estimate the sequencingcoverage of individual libraries as well as to estimatephylogenetic relationships between the three speciesand among the nine T. cacao varieties. These sequencesrepresent single copy genes in Arabidopsis thalianaand tend to be conserved as single copy genes acrossEukaryotes [42]. Since these genes are highly conserved andoften present in a single copy in the genome, they are use-ful for estimating the sequencing coverage of each libraryand estimating copy number of the TE families. The 357putative UCOS homologs in T. cacao were identified usingblastx with an e-value threshold of 1E-34 [43]. The singlecopy status of the UCOS was verified by removing allcontigs that had multiple hits to the T. cacao genome witha e-value lower than 1E-06. This left 245 UCOS to whichSveinsson et al. BMC Genomics 2013, 14:502 Page 9 of 12http://www.biomedcentral.com/1471-2164/14/502the reads were mapped to using BWA, coverage of eachUCOS contig was estimated using bedTools and singlenucleotide polymorphisms (SNPs) called using samtools(see Additional file 5). Finally an average coverage wascalculated for each library, by calculating the meancoverage of the 245 UCOS contigs. Coverage of eachTE reference contig was divided by the mean UCOScoverage, in order to estimate a relative copy-numberof TEs to single copy nuclear genes.Phylogenetic analysis using the UCOS contigsA phylogenetic matrix was constructed by using the 245UCOS contigs identified above as reference for shortread mapping and by calling SNPs using previously de-scribed methods. Theobroma cacao cv. Scavina-6 wasexcluded from the phylogenetic analysis due to low se-quencing coverage. For the construction of the matrix,only positions that were covered by 6 or more high qualityreads in a given sample, with base quality equal or largerto 20 (column 6 in samtools pileup -vcf output) were used.Positions containing any ambiguous nucleotides, i.e. hetero-zygotes, were converted to Ns as were all other positionsthat did not meet previously mentioned criteria. Finally Nswere converted to gaps, trimAl v.1.2 [44] used to removeall gaps and to convert the alignments to nexus format(trimAl flags: -nogap -nexus). All alignments shorter than50 nucleotides were excluded from further analysis, leaving97 UCOS for further analysis. A matrix with positionalinformation of each of the UCOS contigs was constructedusing phyutility v.2.2.4 [45] (phyutility flags: -concat)(Additional file 6), for a combined analysis that includesseparate analyses of each contig using a coalescence-based program (see below). Gene trees of individualUCOS alignments longer than 50 nucleotides were esti-mated with RAxML v7.2.6 [46], using 10 independentruns and the GTRGAMMA sequence substitution model(raxml flags: -m GTRGAMMA -N 10). In order to esti-mate a single phylogeny of the three species and nineremaining T. cacao varieties (Scavina-6 excluded), a STAR(species trees based on average ranks of coalescences)phylogeny [47] was constructed using the phybase R pack-age (v.1.3) [48]. STAR uses the mean ranks of coalescentoccurrences in a set of gene trees to construct a speciestree topology [47]. In order to estimate branch lengthson the STAR tree, model parameters of the entirematrix were estimated using jModelTest v2.0.2 [49,50]and GARLI v2.0 [51] used to optimize model parame-ters and to add branch lengths to the STAR tree. Sup-port values for the STAR phylogeny were estimatedusing a multi locus bootstrap [52] method implementedin the phybase package [48]. One thousand multi locusbootstrap replicates were analyzed using Phyml v3.0[53], STAR trees estimated for each set of bootstrapreplicates and a consensus tree constructed from allthe STAR trees using the consense program in thephylip package v3.69 [54].Graph based clustering of the Illumina readsThe repetitive elements of three species studied here werealso investigated in a de novo fashion using RepeatExplorer[55], which is a graph based clustering method of char-acterizing repetitive elements described in [56], withthe program’s default settings. The Criollo-22 individ-ual was chosen as the T. cacao representative. Briefly,RepeatExplorer uses information from sequence simi-larity among the reads and their partial overlap to con-struct graphs. Graphs are constructed using a Louvainmethod [57], where sequence reads are represented byvertices, edges are connected with overlapping readsand edge weights correspond to the similarity scoreamong reads. The graph layouts are then examined inorder to find separate clusters of reads that are oftenconnected and correspond to distinct families of gen-omic repeats. These clusters are analyzed in regards totheir size, determined by the number of reads compris-ing each cluster as well as their graph topology whichgives information about their structure and variability.RepeatExplorer also performs a sequence similaritysearch of each cluster against RepBase [37] in order toidentify the type of the repetitive elements present inthe cluster. If the predicted number of nodes exceeds thecapacity of the available RAM, RepeatExplorer randomlysubsamples the reads. RepeatExplorer outputs a commaseparated value (csv) file, containing relevant information ofthe clusters it identified and that consist of 0.01% or moreof the reads used in the analysis (default cut-off). The pro-gram calculates the genome percentage, which is the num-ber of reads in each cluster divided by the all the reads usedin the graph based clustering (11,243,224 reads in total). Anin house python script (available on request) was used parsethe csv output file and combine parts of it with the figuresof graph layouts. The output of that script is a panel ofgraph layouts, with each cluster’s most abundant class ofelement, in addition to the genome percentage and numberof paired-end reads belonging to the cluster.Statistical analysisSimilarity of TE composition among sequenced individualswas investigated with a principal component analysis (PCA)using coverage of each TE super-family in the genomes ofH. balaensis, T. grandiflorum and the nine T. cacao var-ieties. The PCA was performed using the prcomp functionin R v2.14.1 [58], using the abundance of each super-familyas explanatory variables with a natural logarithmic trans-formation and scale = TRUE. An in-house R script was usedto run the PCA analysis on all sub-sampled data sets. Thereduced data set ensured that differences in sequencingdepth and read length did not affect the results.Sveinsson et al. BMC Genomics 2013, 14:502 Page 10 of 12http://www.biomedcentral.com/1471-2164/14/502Additional filesAdditional file 1: Relative copy-number of transposable elementsusing reference based mapping to conserved regions of the classI LTR elements. Relative copy-numbers of the TE super-families in thethree species represented with bar plots. Relative copy-number wascalculated by dividing the total coverage of each super-family, within asample, by the sample’s mean UCOS coverage. The mapping waspreformed with relaxed settings in the short read aligner and the readswere mapped to conserved regions of class I LTR elements.Additional file 2: Graph layouts of all the clusters generated in thegraph based clustering analysis. Graph layouts of clusters thatcontained 0.01% or more of the short reads used in the graph basedclustering. Herrania balaensis is shown on the left, T. grandiflorum in themiddle and T. cacao cv. Criollo on the right. Clusters are ordered by size,with largest at the top. Below each graph layout is the class of therepetitive element, the genome percentage of each cluster and numberof paired reads belonging to it in parentheses. Coloured regions in somegraphs represent conserved domains identified by RepeatExplorer. Itshould be noted that a few clusters annotated as “low complexity” mayactually be referable to plastid sequence (e.g. CL19,CL26 & CL78).Additional file 3: Biplots from PCAs on all of the sub-sampleddatasets. Biplots from principal component analysis on everysub-sampled dataset generated, 49 in total. The abundance of each TEsuper-family was used as explanatory variables and the percentage of theexplained variance is shown in parentheses in the legend of the x- andy-axis of the biplot.Additional file 4: The Theobroma cacao TE reference contigs usedin this study. A compressed file in a zip format that contains two textfiles with the reference contigs of the transposable elements used in thisstudy in a fasta format. These contigs were constructed by the authors ofthe T. cacao genome paper [23] who graciously shared them with theauthors of this study.Additional file 5: UCOS contigs used. A text file with containing the245 Theobroma cacao UCOS contigs used in this study. The contigs wereidentified by blasting 357 Arabidopsis thaliana UCOS contigs to theT. cacao genome. One hundred and twelve contigs with more than onesignificant hits (e-value lower that 1E-06) were excluded.Additional file 6: Phylogenetic matrix. The phylogenetic matrix whichwas used to construct Figure 1. The matrix in nexus format that consistsof a concatenated sequence of 97 partially sequenced UCOS genes andthe positional information of individual gene in a Mr.Bayes nexus block.Competing interestsThe authors declared no competing interest concerning the work in thispaper.Authors’ contributionsSS planned the work, performed the data analysis and wrote the manuscript.NG and NCK assisted in data analysis and with writing the manuscript. QCjointly planned the work, co-wrote the manuscript and provided partialfunding for the study. All authors read and approved the final manuscript.AcknowledgementsThe authors would like to thank Armando Geraldes and Charles Hefer foruseful discussion. We are grateful to both Hannes Dempewolf and JanEngels who were instrumental in setting up the cacao sequencing project.We would also like to thank Jeannette Whitton for useful comments andproof reading of the manuscript. We would also like to thank the staff ofCofactor genomics for their helpfulness during sequencing. We are gratefulfor funding for this project received from the World Bank (DevelopmentMarketplace competition), NSERC (discovery grant to Q.C.) and UBCDepartment of Botany scholarship funding (to S.S.).Author details1Department of Botany and Biodiversity Research Centre, University of BritishColumbia, 6270 University Boulevard, Vancouver, BC V6T 1Z4, Canada.2Department of Ecology & Evolutionary Biology, University of ColoradoBoulder, 1800 Colorado Ave, Boulder, CO 80309, USA.Received: 15 April 2013 Accepted: 19 July 2013Published: 24 July 2013References1. Kumar A, Bennetzen JL: Plant retrotransposons. Annu Rev Genet 1999,33:479–532.2. Feschotte C, Pritham EJ: DNA transposons and the evolution of eukaryoticgenomes. Annu Rev Genet 2007, 41:331–368.3. Kelly LJ, Leitch IJ: Exploring giant plant genomes with next-generationsequencing technology. Chromosome Res 2011, 19:1–15.4. Sun C, Shepard DB, Chong RA, Arriaza JL, Hall K, Castoe TA, Feschotte C,Pollock DD, Mueller RL: LTR retrotransposons contribute to genomicgigantism in plethodontid salamanders. Genome Biol Evol 2012, 4:168–183.5. Martin A, Troadec C, Boualem A, Rajab M, Fernandez R, Morin H, Pitrat M,Dogimont C, Bendahmane A: A transposon-induced epigenetic changeleads to sex determination in melon. Nature 2009, 461:1135–1138.6. Zhou L, Mitra R, Atkinson PW, Hickman AB, Dyda F, Craig NL: Transpositionof hAT elements links transposable elements and V (D) J recombination.Nature 2004, 432:995–1001.7. Craig NL, Craigie R, Gellert M, Lambowitz AM: Mobile DNA II. Washington,DC: Amer Society for Microbiology; 2002.8. Boeke JD, Corces VG: Transcription and reverse transcription ofretrotransposons. Annu Rev Microbiol 1989, 43:403–434.9. Wicker T, Sabot F, Hua-Van A, Bennetzen JL, Capy P, Chalhoub B, Flavell A,Leroy P, Morgante M, Panaud O: A unified classification system foreukaryotic transposable elements. Nat Rev Genet 2007, 8:973–982.10. Novick PA, Smith JD, Floumanhaft M, Ray DA, Boissinot S: The evolutionand diversity of DNA transposons in the genome of the lizard Anoliscarolinensis. Genome Biol Evol 2011, 3:1–14.11. Vicient CM, Suoniemi A, Anamthawat-Jónsson K, Tanskanen J, Beharav A,Nevo E, Schulman AH: Retrotransposon BARE-1 and its role in genomeevolution in the genus Hordeum. Plant Cell Online 1999, 11:1769–1784.12. Pearce SR, Knox M, Ellis TH, Flavell AJ, Kumar A: Pea Ty1-copia groupretrotransposons: transpositional activity and use as markers to studygenetic diversity in Pisum. Mol Gen Genet 2000, 263:898–907.13. Huang X, Lu G, Zhao Q, Liu X, Han B: Genome-wide analysis of transposoninsertion polymorphisms reveals intraspecific variation in cultivated rice.Plant Physiol 2008, 148:25–40.14. Kumar A, Hirochika H: Applications of retrotransposons as genetic tools inplant biology. Trends Plant Sci 2001, 6:127–134.15. Syed N, Sureshsundar S, Wilkinson M, Bhau B, Cavalcanti J, Flavell A:Ty1-copia retrotransposon-based SSAP marker development in cashew(Anacardium occidentale L.). Theor Appl Genet 2005, 110:1195–1202.16. Schulman AH, Flavell AJ, Paux E, Ellis T: The application of LTR retrotransposonsas molecular markers in plants. Methods Mol Biol 2012, 859:115–153.17. Flavell AJ, Smith DB, Kumar A: Extreme heterogeneity of Ty1-copia groupretrotransposons in plants. Mol Gen Genet 1992, 231:233–242.18. Wood GAR, Lass RA: Cocoa. 4th edition. Blackwell, UK: Longman Group; 2001.19. Motamayor JC, Lachenaud P, e Mota JWS, Loor R, Kuhn DN, Brown JS,Schnell RJ: Geographic and genetic population differentiation of theAmazonian chocolate tree (Theobroma cacao L). PLoS One 2008, 3:e3311.20. Cheesman EE: Notes on the nomenclature, classification and possiblerelationships of cocoa populations. Trop Agric 1944, 21:144–159.21. Motamayor JC, Risterucci AM, Heath M, Lanaud C: Cacao domestication II:progenitor germplasm of the Trinitario cacao cultivar. Heredity 2003,91:322–330.22. Kane N, Sveinsson S, Dempewolf H, Yang JY, Zhang D, Engels JMM, Cronk Q:Ultra-barcoding in cacao (Theobroma spp.; Malvaceae) using wholechloroplast genomes and nuclear ribosomal DNA. Am J Bot 2012, 99:320–329.23. Argout X, Salse J, Aury JM, Guiltinan MJ, Droc G, Gouzy J, Allegre M,Chaparro C, Legavre T, Maximova SN: The genome of Theobroma cacao.Nat Genet 2010, 43:101–108.24. Chaparro C, Sabot F: Methods and software in NGS for TE analysis.Methods Mol Biol 2012, 859:105–114.25. Tenaillon MI, Hufford MB, Gaut BS, Ross-Ibarra J: Genome size andtransposable element content as determined by high-throughputsequencing in maize and Zea luxurians. Genome Biol Evol 2011, 3:219.Sveinsson et al. BMC Genomics 2013, 14:502 Page 11 of 12http://www.biomedcentral.com/1471-2164/14/50226. Sabot F, Picault N, El-Baidouri M, Llauro C, Chaparro C, Piegu B, Roulin A,Guiderdoni E, Delabastide M, McCombie R: Transpositional landscape ofthe rice genome revealed by paired-end mapping of high-throughputre-sequencing data. Plant J 2011, 66:241–246.27. Macas J, Neumann P, Navratilova A: Repetitive DNA in the pea(Pisum sativum L.) genome: comprehensive characterization using 454sequencing and comparison to soybean and Medicago truncatula.BMC Genomics 2007, 8:427.28. Wicker T, Narechania A, Sabot F, Stein J, Vu G, Graner A, Ware D, Stein N:Low-pass shotgun sequencing of the barley genome facilitates rapididentification of genes, conserved non-coding sequences and novelrepeats. BMC Genomics 2008, 9:518.29. Hribova E, Neumann P, Matsumoto T, Roux N, Macas J, Doležel J: Repetitivepart of the banana (Musa acuminata) genome investigated by low-depth 454 sequencing. BMC Plant Biol 2010, 10:204.30. Wessler SR: Transposable elements and the evolution of eukaryoticgenomes. Proc Natl Acad Sci 2006, 103:17600–17601.31. Gabriel A, Willems M, Mules EH, Boeke JD: Replication infidelity during asingle cycle of Ty1 retrotransposition. Proc Natl Acad Sci 1996, 93:7767–7771.32. Kalendar R, Tanskanen J, Immonen S, Nevo E, Schulman AH: Genomeevolution of wild barley (Hordeum spontaneum) by BARE-1retrotransposon dynamics in response to sharp microclimaticdivergence. Proc Natl Acad Sci 2000, 97:6603–6607.33. Figueira A, Janick J, Goldsbrough P: Genome size and DNA polymorphismin Theobroma cacao. J Am Soc Hortic Sci 1992, 117:673–677.34. Marie D, Brown SC: A cytometric exercise in plant DNA histograms, with2C values for 70 species. Biological Cell 1993, 78:41–51.35. Xu Z, Wang H: LTR_FINDER: an efficient tool for the prediction of full-lengthLTR retrotransposons. Nucleic Acids Res 2007, 35:W265–W268.36. Ellinghaus D, Kurtz S, Willhoeft U: LTRharvest, an efficient and flexiblesoftware for de novo detection of LTR retrotransposons. BMC Bioinforma2008, 9:18.37. Jurka J, Kapitonov VV, Pavlicek A, Klonowski P, Kohany O, Walichiewicz J:Repbase update, a database of eukaryotic repetitive elements. CytogenetGenome Res 2005, 110:462–467.38. Li H, Durbin R: Fast and accurate short read alignment with Burrows–Wheelertransform. Bioinformatics 2009, 25:1754–1760.39. Quinlan AR, Hall IM: BEDTools: a flexible suite of utilities for comparinggenomic features. Bioinformatics 2010, 26:841–842.40. Li H, Handsaker B, Wysoker A, Fennell T, Ruan J, Homer N, Marth G, AbecasisG, Durbin R: The sequence alignment/map format and SAMtools.Bioinformatics 2009, 25:2078–2079.41. Steinbiss S, Willhoeft U, Gremme G, Kurtz S: Fine-grained annotation andclassification of de novo predicted LTR retrotransposons. Nucleic Acids Res2009, 37:7002–7013.42. Kozik A, Matvienko M, Kozik I, Van Leeuwen H, Van Deynze A, Michelmore R:Eukaryotic ultra conserved orthologs and estimation of gene capture InEST libraries [abstract]. Plant and Animal Genomes Conference 2008, 16:P6.43. Altschul SF, Madden TL, Schäffer AA, Zhang J, Zhang Z, Miller W, Lipman DJ:Gapped BLAST and PSI-BLAST: a new generation of protein databasesearch programs. Nucleic Acids Res 1997, 25:3389–3402.44. Capella-Gutierrez S, Silla-Martinez JM, Gabaldon T: trimAl: a tool forautomated alignment trimming in large-scale phylogenetic analyses.Bioinformatics 2009, 25:1972–1973.45. Smith SA, Dunn CW: Phyutility: a phyloinformatics tool for trees,alignments and molecular data. Bioinformatics 2008, 24:715–716.46. Stamatakis A: RAxML-VI-HPC: maximum likelihood-based phylogeneticanalyses with thousands of taxa and mixed models. Bioinformatics 2006,22:2688–2690.47. Liu L, Yu L, Pearl DK, Edwards SV: Estimating species phylogenies usingcoalescence times among sequences. Syst Biol 2009, 58:468–477.48. Liu L, Yu L: PHYBASE: an R package for phylogenetic analysis.Bioinformatics 2010, 26:962–963.49. Guindon S, Gascuel O: A simple, fast, and accurate algorithm to estimatelarge phylogenies by maximum likelihood. Syst Biol 2003, 52:696–704.50. Posada D: jModelTest: phylogenetic model averaging. Mol Biol Evol 2008,25:1253–1256.51. Zwickl DJ: Genetic algorithm approaches for the phylogenetic analysis of largebiological sequence datasets under the maximum likelihood criterion. PhDthesis. University of Texas; 2006.52. Seo TK: Calculating bootstrap probabilities of phylogeny using multilocussequence data. Mol Biol Evol 2008, 25:960–971.53. Guindon S, Dufayard JF, Lefort V, Anisimova M, Hordijk W, Gascuel O: Newalgorithms and methods to estimate maximum-likelihood phylogenies:assessing the performance of PhyML 3.0. Syst Biol 2010, 59:307–21.54. Felsenstein J: PHYLIP (Phylogeny Inference Package) version 3.6. Distributed bythe author. Seattle: Department of Genetics, University of Washington; 1993.55. Novak P, Neumann P, Pech J, Steinhaisl J, Macas J: RepeatExplorer:a Galaxy-based web server for genome-wide characterization ofeukaryotic repetitive elements from next generation sequence reads.Bioinformatics 2013, 29:792–793.56. Novák P, Neumann P, Macas J: Graph-based clustering andcharacterization of repetitive sequences in next-generation sequencingdata. BMC Bioinforma 2010, 11:378.57. Blondel VD, Guillaume JL, Lambiotte R, Lefebvre E: Fast unfolding ofcommunities in large networks. Journal of Statistical Mechanics: Theory andExperiment 2008, 200(8), P10008.58. Ihaka R, Gentleman R: R: a language for data analysis and graphics.J Comp Graph Stat 1996, 5:299–314.doi:10.1186/1471-2164-14-502Cite this article as: Sveinsson et al.: Transposon fingerprinting using lowcoverage whole genome shotgun sequencing in Cacao (Theobromacacao L.) and related species. BMC Genomics 2013 14:502.Submit your next manuscript to BioMed Centraland take full advantage of: • Convenient online submission• Thorough peer review• No space constraints or color figure charges• Immediate publication on acceptance• Inclusion in PubMed, CAS, Scopus and Google Scholar• Research which is freely available for redistributionSubmit your manuscript at www.biomedcentral.com/submitSveinsson et al. BMC Genomics 2013, 14:502 Page 12 of 12http://www.biomedcentral.com/1471-2164/14/502


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items