Accurate prediction of binding affinities for charged molecules remains a challenge in molecular simulation. The modelling of protonation and electronic polarization changes upon ligand transfer from water to the protein interior are routinely unaccounted for and the contributions of these processes are still unclear. Here we benchmark the predictive performance of absolute alchemical binding free energy methods on 10 small molecule inhibitors targeting urokinase plasminogen activator, a serine protease regularly found overexpressed in metastatic tumors. Our ligand set includes highly charged, flexible, and titratable molecules that present a challenging case study evaluating the difficulties of modeling electrostatic and polarization effects in protein-ligand binding interactions. We extensively sample each inhibitor with ensemble MD trajectories totaling ~340 µs to compare the impact of simulation conditions including: counter-ion removal to maintain system charge neutrality, ligand restraint strategies to improve convergence, increasing salt concentration for electrostatic screening, and varying binding site protonation states. We further demonstrate an alternative approach based on implicit PBSA solvation analysis that conveniently accounts for the impact of polarization and improves prediction accuracy through scaling of the protein dielectric constant. Our results provide a set of best-practices for the application of absolute alchemical binding free energy predictions that will accelerate the drug screening process and enables consideration of electrostatic and polarization effects.

Electrostatics and polarization in protein-ligand binding affinity predictions

ACS Spring Meeting 2021

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.