Rapid detection of identity-by-descent tracts for mega-scale datasets

Abstract

The ability to identify segments of genomes identical-by-descent (IBD) is a part of standard workflows in both statistical and population genetics. However, traditional methods for finding local IBD across all pairs of individuals scale poorly leading to a lack of adoption in very large-scale datasets. Here, we present iLASH, IBD by LocAlity-Sensitive Hashing, an algorithm based on similarity detection techniques that shows equal or improved accuracy in simulations compared to the current leading method and speeds up analysis by several orders of magnitude on genomic datasets, making IBD estimation tractable for hundreds of thousands to millions of individuals. We applied iLASH to the Population Architecture using Genomics and Epidemiology (PAGE) dataset of ~52,000 multi-ethnic participants, including several founder populations with elevated IBD sharing, which identified IBD segments on a single machine in an hour (~3 minutes per chromosome compared to over 6 days per chromosome for a state-of-the-art algorithm). iLASH is able to efficiently estimate IBD tracts in very large-scale datasets, as demonstrated via IBD estimation across the entire UK Biobank (~500,000 individuals), detecting nearly 13 billion pairwise IBD tracts shared between ~11% of participants. In summary, iLASH enables fast and accurate detection of IBD, an upstream step in applications of IBD for population genetics and trait mapping.Inferring segments of the genome inherited Identical-By-Descent (IBD) is a standard method in modern genomics pipelines to understand population structure and infer relatedness across datasets 1-6 . Furthermore, it can be leveraged for alternative mapping strategies such as populationbased linkage 7 , capturing rare variation from array datasets 8 , and improving long-range phasing. However, the ability to scale this process to mega-scale datasets while comparing individuals along the genome has been limited. While approximate methods have been developed to improve

Identity-by-descent detection across 487,409 British samples reveals fine scale population structure and ultra-rare variant associations

“…Our analysis detected an average of 1.8 IBD segments per pair in the UK Biobank dataset within the past 50 generations. This is consistent with a previous study focusing on longer and more recent segments (average of 0.1 segments >2.9 cM per pair 66 ), but less than another recent study in a similar length range (average 1.96 segments >2 cM per pair 67 ). Taking uncertainty of the detected IBD segments into account may reconcile these estimates.…”

“…FastSMC's identification step currently relies on the GERMLINE2 genotype hashing strategy. It will be interesting to test other heuristic strategies for rapidly identifying identical segments, such as the locality-sensitive hashing strategy recently implemented in the iLASH algorithm (exhibiting 95% concordance with GERMLINE in application to real multi-ethnic data 66 ), or methods that rely on the positional Burrows-Wheeler transform (PBWT) data structure 17,67,68 . Several methods now exist to reconstruct gene genealogies in large samples [69][70][71][72] .…”

A rapid, accurate approach to inferring pedigrees in endogamous populations

“…PONDEROSA takes as input IBD segment estimates (either in GERMLINE or iLASH format), as well as pairwise IBD1 and IBD2 values in a KING-formatted file (to define parent-offspring pairs) and a PLINK-formatted .fam file (to define known paths through the pedigree) (Gusev et al 2009;Shemirani and Belbin 2019;Manichaikul et al 2010). The user can also supply reported age data (to constrict possible pedigree relationships) and a PLINK .ped file (to more accurately merge IBD segments).…”

RAFFI: Accurate and fast familial relationship inference in large scale biobank studies using RaPID

“…This is because the former has a linear time algorithm [ 19 ] while the latter needs quadratic time algorithms. Among fast IBD segment detection methods, hash table-based methods [ 16 , 17 ] are typically memory intensive. RaPID [ 15 ] and hap-IBD [ 18 ] are based on the scanning algorithm of PBWT and are scaling up both in terms of run time and memory.…”