SV2: Accurate Structural Variation Genotyping and De Novo Mutation Detection

Structural Variation (SV) detection from short-read Illumina whole genome sequencing is error prone, presenting significant challenges for analysis in particularly, de novo mutations. Here we describe SV2, a machine-learning algorithm for genotyping deletions and tandem duplications from paired-end whole genome sequencing data. SV2 can rapidly integrate variant calls from multiple structural variant discovery algorithms into a unified callset with low rates of false discoveries and Mendelian errors, accurate de novo detection with no transmission bias in families.