Assessment of Different Parameters on the Accuracy of Computational Alanine Scanning of Protein–Protein Complexes with the Molecular Mechanics/Generalized Born Surface Area Method

Computational
alanine scanning with the molecular mechanics
generalized
Born surface area (MM/GBSA) method constitutes a widely used approach
for identifying critical residues at protein–protein interfaces.
Despite its popularity, the MM/GBSA method still has certain drawbacks
due to its dependence on many factors. Here, we performed a systematical
study on the impact of four different parameters, namely, the internal
dielectric constant, the generalized Born model, the entropic term,
and the inclusion of structural waters on the accuracy of computational
alanine scanning calculations with the MM/GBSA method. Our results
show that the internal dielectric constant is the most critical parameter
for getting accurate predictions. The introduction of entropy and
interfacial water molecules decreased the quality of the predictions,
while the generalized Born model had little to no effect. Considering
the significance of the internal dielectric value, we proposed a methodology
based on the energetic predominance of a particular set of amino acids
at the protein–protein interface for selecting an appropriate
value for this variable. We hope that these results serve as a guideline
for future studies of protein–protein complexes using the MM/GBSA
method.