Genome-wide selection inference at short tandem repeats

Short tandem repeats (STRs) comprising repeated sequences of 1-6 bp are one of the largest sources of genetic variation in humans. STRs are known to contribute to a variety of disorders, including Mendelian diseases, complex traits, and cancer. Based on their functional importance, mutations at some STRs are likely to introduce negative effects on reproductive fitness over evolutionary time. We previously developed SISTR (Selection Inference at STRs), a population genetics framework to measure negative selection against individual STR alleles. Here, we extend SISTR to enable joint estimation of the distribution of selection coefficients across a set of STRs. This method (SISTR2) allows for more accurate analysis of a broader range of STRs, including loci with low mutation rates. We apply SISTR2 to explore the range of feasible mutation parameters and demonstrate substantial variation in mutation and selection parameters across different classes of STRs. Finally, we estimate the relative burden of de novo and inherited variation at STR vs. single nucleotide variants (SNVs). Our results suggest that whereas SNVs contribute a greater total burden of inherited variation in a typical genome, the burden of de novo mutations at STRs is greater than that of SNVs. Overall, we anticipate that the evolutionary insights gained from this study will be important for future studies of variation at STRs and their role in evolution and disease.