Amino acid exchangeability and surface accessibility underpin the effects of single substitutions

Abstract

Deep mutational scans have measured the effects of many mutations on many different proteins. Here we use a collection of such scans to perform a statistical meta-analysis of the effects of single amino acid substitutions. Specifically, we model the relative deleteriousness of each substitution in each deep mutational scan with respect to the identities of the wildtype and mutant residues, and the wildtype residues surface accessibility. This model explains much of the variance in mutational effects and quantifies physicochemical trends underlying them, including by yielding an empirical amino acid exchangeability matrix.