Computational prediction of binding hotspots

We combine side-chain modeling, energy minimization and binding free energy calculation to predict point mutations with significant impacts on binding affinities (binding hotspots). Our method achieves high accuracy for two datasets (alanine-scanning mutations in ASEdb and 570 mutations on protease-inhibitor complexes). In particular, we can predict mutations that lead to improved binding with success. We discuss various factors that may contribute the prediction accuracy, including the amino acid to mutate to, and the position of the mutation.