Methods for accurate quantification of LTR-retrotransposon copy number using short-read sequence data: a case study in Sorghum

Abstract

Transposable elements (TEs) are ubiquitous in eukaryotic genomes and their mobility impacts genome structure and function in myriad ways. Because of their abundance, activity, and repetitive nature, the characterization and analysis of TEs remain challenging, particularly from short-read sequencing projects. To overcome this difficulty, we have developed a method that estimates TE copy number from short-read sequences. To test the accuracy of our method, we first performed an in silico analysis of the reference Sorghum bicolor genome, using both reference-based and de novo approaches. The resulting TE copy number estimates were strikingly similar to the annotated numbers. We then tested our method on real short-read data by estimating TE copy numbers in several accessions of S. bicolor and its close relative S. propinquum. Both methods effectively identify and rank similar TE families from highest to lowest abundance. We found that de novo characterization was effective at capturing qualitative variation, but underestimated the abundance of some TE families, specifically families of more ancient origin. Also, interspecific reference-based mapping of S. propinquum reads to the S. bicolor database failed to fully describe TE content in S. propinquum, indicative of recent TE activity leading to changes in the respective repetitive landscapes over very short evolutionary timescales. We conclude that reference-based analyses are best suited for within-species comparisons, while de novo approaches are more reliable for evolutionarily distant comparisons.

Retrotransposons in Plant Genomes: Structure, Identification, and Classification through Bioinformatics and Machine Learning

“…Interestingly, most of the software used to identify TEs requires assembled sequences as input, even though assembly algorithms have trouble with highly repetitive sections of genomes [4,66,203,204]. Repeats cause branches in graphs used in assembly algorithms (which can be one of two classes: overlap-based and De Bruijn graph) [205], leading assemblers to create false joins and wrong copy numbers, or even break graphs at these branch points, generating an accurate but fragmented assembly [205].…”

Evolutionary Dynamics of Transposable Elements Following a Shared Polyploidization Event in the Tribe Andropogoneae

“…reporting the number of reads per cluster, which can then be used to estimate the genome space occupied by each particular repeat, i.e., (total length of each cluster (in Mb) x genome size (in Mb)) / total length of all clusters (in Mb) (Macas et al 2015, Kelly et al 2015, Ramachandran et al 2016. For species-specific clustering, three million reads (approximately 0.2x to 0.5x genome coverage) were sub-sampled from each dataset and processed to the format required by RepeatExplorer.…”

“…The maize TE annotation file (B73v4.TE.filtered.gff3) was downloaded from Gramene Release 62 (http://www.gramene.org) and sorghum retroelement annotations were retrieved from our previous work described in Ramachandran et al 2016. BEDtools closest (version 2.17.0; Quinlan and Hall 2010) was used to identify LTR-retrotransposon (gypsy and copia) insertions within 1-5 kb upstream of genes.…”

Evolutionary dynamics of transposable elements following a shared polyploidization event in the tribe Andropogoneae

“…reporting the number of reads per cluster, which can then be used to estimate the genome space occupied by each particular repeat, i.e., (total length of each cluster (in Mb) x genome size (in Mb)) / total length of all clusters (in Mb) (Kelly et al 2015, Ramachandran et al 2016. For species-specific clustering, three million reads (approximately 0.2x to 0.5x genome coverage) were sub-sampled from each dataset and processed to the format required by RepeatExplorer.…”