Missense3D: leveraging protein structures, including AlphaFold models, to understand the effect of missense variants.

Abstract

The growing availability of 3D protein structures enables large-scale assessment of missense variants and their phenotypic impact. Missense3D, first developed in 2019, predicts the structural consequences of amino acid substitutions and provides mechanistic explanations by identifying features such as cavities, steric clashes and disrupted non-covalent bonds that may cause protein misfolding. The new Missense3D algorithm presented here accepts genomics, protein, and structural coordinates; models variants on both single proteins and protein-protein complexes using experimental and predicted structures, including AlphaFold models; performs automated sequence-structure mapping; and evaluates structural model quality before analysis. The Missense3D web server is freely available to all users, including commercial ones, at https://missense3d.bc.ic.ac.uk/~missense3d2/.