Inference of elevated mutation rates and variant effects using 700k exomes

Abstract

Genomic sequencing is now widely accessible for genetic diagnostics and is emerging as a component of newborn screening. This technological development generates the need to characterize incoming mutations, create comprehensive datasets of genes causing rare Mendelian disorders, and identify pathogenic variants. Large-scale exome sequencing datasets such as Genome Aggregation Database (gnomAD) have been assembled to help address these challenges. The recent release of gnomAD (v4; n = 730,947) uncovers millions of rare coding variants, many of which have arisen more than once by independent recurrent mutations in the rapidly growing recent human population. Here, we use newly developed theoretical understanding of sampling properties of rare variants to estimate key population genetics parameters of practical importance to human genetics such as demography history, mutation rate, and selection. Solely relying on population data, our method Population Inferred Estimates of Selection (PIES) identifies novel genes with loss-of-function mutational hotspots likely due to selection in spermatogonia. PIES efficiently estimates selection coefficients for heterozygous loss-of-function variants. Combining population genetics inference with variant effect predictors, PIES predicts pathogenic missense mutations and improves variant prioritization for genetic diagnostics and newborn screening.