Assessing Foldx for Predicting Protein-Protein Binding Affinity Changes Due to Multiple Mutations

Abstract

Point mutations can affect the binding affinity between two proteins. Multiple point mutations can interact, affecting binding in a non-additive way (epistasis). The ability of current methods to predict epistatic phenomena has not been thoroughly explored. FoldX, a widely used semi-empirical forcefield based software suite allows fast prediction of single mutation effects. It is therefore of interest to how it performs with higher order mutants. Here we have tested the performance of FoldX for predicting binding affinity changes and possible epistatic effects in protein-protein complexes due to multiple mutations. We obtained experimental data for test systems from the SKEMPI v2.0 database, which contains 572 double mutant records with constituent singles. Corresponding pdb files were acquired from the RCSB and pre-processed using PDBFixer to correct any errors. FoldX was then used for structure minimization and calculation of free energy of binding for the wild type and mutants. Accuracy enhancements for FoldX were also tested, including using snapshots generated via molecular dynamics as inputs to provide conformational flexibility. From this work we expect to develop a better understanding of what factors may indicate epistatic phenomena and how to enhance or develop methods to better predict these effects in protein systems.