Population Genetic Inference with Identity by Descent

Abstract

Identical-by-descent (IBD) segments carry important information on genetic relatedness between individuals and therefore are used widely for statistical inference in population genetics. In this work, we focus on the inference of the genome-wide mutation rate using IBD segments. The two primary methods for estimating the genome-wide mutation rate have been counting de novo mutations in parent-offspring trios and comparing sequence data between closely related species. With parent-offspring trio analysis it is difficult to control for genotype error, and resolution is limited because each trio provides information from only two meioses. Inter-species comparison is difficult to calibrate due to uncertainty in the number of meioses separating species, and it can be biased by selection and by changing mutation rates over time. Existing IBD-based methods are limited to highly inbred samples, or lack robustness to genotype error and error in the estimated demographic history. In this work, we propose mathematical models based on coalescent theory for an IBD-based method that uses sharing of IBD segments among sets of three individuals to estimate the mutation rate as well as the gene conversion rate. With the application to whole genome sequence data from TOPMed Project, we obtain an estimate of 1.58E-8 mutations per base pair per meiosis based on European descendants and 1.52E-8 based on African descendants.