Population-specific recombination maps from segments of identity by descent

Recombination rates vary significantly across the genome, and estimates of recombination rates are needed for downstream analyses such as haplotype phasing and genotype imputation. Existing methods for recombination rate estimation are limited by insufficient amounts of informative genetic data or by high computational cost. We present a method for using segments of identity by descent to infer recombination rates. Our method can be applied to sequenced population cohorts to obtain high-resolution, population-specific recombination maps. We use our method to generate new recombination maps for European Americans and for African Americans from TOPMed sequence data from the Framingham Heart Study (1626 unrelated individuals) and the Jackson Heart Study (2046 unrelated individuals). We compare our maps to existing maps using the Pearson correlation between estimated recombination rates. In Europeans we use the deCODE map, which is based on a very large set of Icelandic family data (126,407 meioses), as a gold standard against which to compare other maps. Our European American map has higher accuracy at fine-scale resolution (1-10kb) than linkage disequilibrium maps from the HapMap and 1000 Genomes projects. Our African American map has much higher accuracy than an admixture-based map that is derived from a similar number individuals, and similar accuracy at fine scales (1-10kb) to an admixture-based map that is derived from 15 times as many individuals.