Evaluation and application of MD-PB/SA in structure-based hierarchical virtual screening.

Molecular dynamics (MD) based molecular mechanics Poisson-Boltzmann and surface area (MM-PB/SA) calculation (MD-PB/SA) has been widely used to estimate binding free energies for receptor-ligand complexes. While numerous reports have focused on assessing accuracy and efficiency, fewer studies have paid attention to performance in lead discovery. In the present study, we report a critical evaluation of MD-PB/SA in hierarchical virtual screening (HVS) both theoretically and practically. It is shown that based on native poses, MD-PB/SA could be well applied to predict the relative binding energy for both congeneric and diverse ligands for different protein targets. However, there is a limitation for MD-PB/SA to distinguish the native pose of one ligand from the artificial pose of another when a huge difference exists between two molecules. By combining a physics-based scoring function with a knowledge-based structural filter, we improve the predictability and validate the practical use of MD-PB/SA in lead discovery by identifying novel inhibitors of p38 MAP kinase. We also expand our study to other protein targets such as HIV-1 RT and NA to assess the general validity of MD-PB/SA.