Integrated genotyping of structural variation

Abstract

Discovering genotype of structural variations (SV) is a new and challenging topic. To the best of our knowledge, estimation of variant allele frequency (VAF) of an SV from both read depth and read alignment has not been done. In this study, we propose BreakDown, a new statistical model that integrates read depth, discordant and split paired-end read alignment to accurately estimate SVs' genotypes and VAFs. Application to 43 low coverage samples in the 1000 Genomes Project shows that BreakDown produces highly accurate genotypes, especially on heterozygous deletions. Estimation of VAFs on both low- and high-coverage data is shown to converge on expected values with small divergence.