A scalable approach for genome-wide inference of ancestral recombination graphs

Abstract

The ancestral recombination graph (ARG) is a graph-like structure that encodes a detailed genealogical history of a set of individuals along the genome. ARGs that are accurately reconstructed from genomic data have several downstream applications, but inference from data sets comprising millions of samples and variants remains computationally challenging. We introduce Threads, a threading-based method that significantly reduces the computational costs of ARG inference while retaining high accuracy. We apply Threads to infer the ARG of 487,409 genomes from the UK Biobank using ~10 million high-quality imputed variants, reconstructing a detailed genealogical history of the samples while compressing the input genotype data. Additionally, we develop ARG-based imputation strategies that increase genotype imputation accuracy for ultra-rare variants (MAC ≤ 10) from UK Biobank exome sequencing data by 5-10%. We leverage ARGs inferred by Threads to detect associations with 52 quantitative traits in non-European UK Biobank samples, identifying 22.5% more signals than ARG-Needle. These analyses underscore the value of using computationally efficient genealogical modeling to improve and complement genotype imputation in large-scale genomic studies.