High-resolution population-specific recombination rates and their effect on phasing and genotype imputation

Abstract

Previous research has shown that using population-specific reference panels has a significant effect on downstream population genomic analyses like haplotype phasing, genotype imputation, and association, especially in the context of population isolates. Here, we developed a high-resolution recombination rate mapping at 10 and 50 kb scale using high-coverage (20–30×) whole-genome sequenced data of 55 family trios from Finland and compared it to recombination rates of non-Finnish Europeans (NFE). We tested the downstream effects of the population-specific recombination rates in statistical phasing and genotype imputation in Finns as compared to the same analyses performed by using the NFE-based recombination rates. We found that Finnish recombination rates have a moderately high correlation (Spearman’s ρ = 0.67–0.79) with NFE, although on average (across all autosomal chromosomes), Finnish rates (2.268 ± 0.4209 cM/Mb) are 12–14% lower than NFE (2.641 ± 0.5032 cM/Mb). Finnish recombination map was found to have no significant effect in haplotype phasing accuracy (switch error rates ~2%) and average imputation concordance rates (97–98% for common, 92–96% for low frequency and 78–90% for rare variants). Our results suggest that haplotype phasing and genotype imputation mostly depend on population-specific contexts like appropriate reference panels and their sample size, but not on population-specific recombination maps. Even though recombination rate estimates had some differences between the Finnish and NFE populations, haplotyping and imputation had not been noticeably affected by the recombination map used. Therefore, the currently available HapMap recombination maps seem robust for population-specific phasing and imputation pipelines, even in the context of relatively isolated populations like Finland.

Sex-biased admixture and assortative mating shape genetic variation and influence demographic inference in admixed Cabo Verdeans

“…In this case, common ancestors within the past 20 generations (post-admixture in Cabo Verde) would give an expected IBD segment length of approximately 2.5 Mb or longer, so we applied 2.5 Mb as the shorter/long ROH boundary. This approximation uses the commonly applied baseline recombination rate of 1 cM/Mb, though we note that recombination varies among populations and along genomes (Hassan et al 2021). Under these two classification schemes, we examined the distributions of total ROH per genome (sum of ROH segments of a specific length) within Cabo Verde, partitioning the distributions into length classes that are enriched for pre- and post-colonization processes.…”

Genetic associations of protein-coding variants in human disease

“…Finland is a well-established genetic isolate and a unique gene pool distinguishes Finns from other Europeans 7 . The distinct Finnish haplotype structure is characterized by large blocks of co-inherited DNA in linkage disequilibrium and an enrichment for alleles that are rare in other populations, but can still be confidently imputed from genotyping data even in the rare and ultra-rare allele frequency spectrum [8][9][10] . Through combining imputed genotypes with detailed phenotypes ascertained through national registries, FG holds the promise to provide particular insights into the so far little examined allele frequency spectrum between 0.1 and 2%, where both sequencing studies and GWAS have so far remained largely underpowered in relation to identifying associations with disease.…”

The Recombination Landscape of the Khoe-San - the Upper Limits of Recombination Divergence in Humans

“…3 and 4). However, it should be noted that recent studies found that population-specific recombination maps have little effect on phasing (Hassan et al 2020), imputation (Hassan et al 2020) and local ancestry inference (van Eeden, Uren, van der Spuy, et al 2021). Therefore, the combined Phase II HapMap recombination map's proxy status with regards to the Nama is dependent on the analysis that the map is used for.…”