Fig. 1: Detecting runs of homozygosity using a reference panel.

a Illustration of genotype data for a diploid individual. b Mapping sequencing reads to a biallelic SNP produces counts of reads for each allele, from which in turn pseudo-haplotype genotypes, i.e., single reads per site, are sampled. c Schematic of Method. A target individual’s genotype data is modeled as mosaic copied from haplotypes in a reference panel (ROH states, colored) and one additional background state (non-ROH, gray). d We applied our method to simulated data with known ROH copied in (see Supplementary Note 1.7 for details). We copied 500 ROH of exactly 4, 6, 8, and 10 cM into 100 artificial chromosomes, and depict histograms of inferred ROH lengths (in color) as well as false positives (in gray) after downsampling and adding errors (0.5× of all 1240K SNPs, with 1% error added). e Down-sampling experiment of a high coverage ancient individual who lived 45,000 years ago (Ust Ishim man) using a modern reference panel (1000 Genomes dataset). We down-sampled pseudohaploid data at the 1240K SNPs. For each target coverage, we ran 100 independent replicates and depict the mean and standard deviation of the inferred ROH in four length bins (4–8, 8–12, 12–20, and >20 cM). The horizontal lines indicate high confidence estimates using diploid genotype calls from all available data.