Accurate prediction of binding affinities for highly charged protein-protein interactions remains a significant computational challenge, particularly when using the widely employed molecular mechanics Poisson-Boltzmann surface area (MM-PBSA) method. Classical force fields with fixed partial charges often overestimate electrostatic contributions in such systems, leading to quantitatively incorrect binding energy predictions. Here, we present a systematic evaluation of critical methodological factors affecting MM-PBSA calculations for highly charged protein-protein complexes using the biologically relevant NDR2/LATS1-MOB1A system from the Hippo signaling pathway as a model. Through extensive molecular dynamics simulations and binding free-energy calculations, we demonstrate that the multi-trajectory approach significantly outperforms the single trajectory method, as the multi-trajectory approach allows receptors and ligands to relax independently. This approach is better able to capture the key conformational changes that occur upon binding. Water model selection proves equally critical and could reverse the predicted binding strength due to inadequate conformational sampling and altered hydration structure around protein surfaces. Furthermore, it is important to note that different ionic concentrations could significantly change the calculated MM-PBSA result. Since the protein system is neutralized by counterions, the electrostatic field of the protein can still interact with its periodic images, creating unphysical contributions to the energetics. Thus, the treatment of various ionic concentrations could have a strong influence on the predicted binding affinity. The findings from this study are hoped to provide practical guidelines for computational researchers working on highly charged protein-protein interactions.
Lin et al. (Sun,) studied this question.