Detection of short identity by descent segments using low-frequency variants

Abstract

AbstractIdentity by descent (IBD) segments are inherited genomic sequences from the same common ancestors that serve as biomarkers during lineage analysis, disease mutation mapping, and within broader fields of population genetics. Current computational tools used for the detection of IBD segments focus on identifying long IBD segments, with short IBD segments (i.e. shorter than 2 cM) often going undetected. Here, we present SILO, a method with a remarkably improved ability to detect short IBD segments as compared to state-of-the-art IBD detection tools. SILO detects IBD segments by considering both common and low-frequency variants (LFVs), where the probability of the presence of a LFV minor allele is assumed to follow a Bernoulli distribution with a Beta prior. SILO was benchmarked against GERMLINE2, hap-IBD, HapFABIA, Parente, and TRUFFLE on simulated data and on a real pedigree from the 1,000 Genomes Project. The results show that SILO has an unrivaled ability to detect short IBD segments, while its ability to detect long IBD segments is similar to that of contemporary tools in the field.