Characterizing Structural Variation in Genomes (from humans to crops)

Several bioinformatics methods have been proposed for the detection and characterization of genomic structural variation from ultra-high throughput genome resequencing data. Although some of these methods demonstrate reasonably high specificity, the sensitivity of available approaches is rather low. We propose a novel method for the identification of genomic structural variation from high throughput paired end genome resequencing data. While utilizing deviations from expected library insert sizes, our approach employs additional information from local patterns of read mapping and supervised learning to predict the position and nature of structural variants. We show that our method shows notably increased sensitivity at no cost in specificity with respect to existing insert size-based tools in the identification of structural variants in the human genome. Furthermore, we show that the additional information incorporated in our approach allow us to make reliable predictions of very short insertions and deletions that are otherwise only recovered by approaches based on the split mapping of resequencing reads.