Consider Solvation and Entropy in the Knowledge-Based Scoring Function for Protein−Ligand Interactions

Abstract

One of the challenges in the development of knowledge-based scoring functions is how to include the solvation and entropic effects. Here, we have presented a computational model to account for the solvation effect, the ligand conformational entropy and the ligand vibrational entropy in the knowledge-based scoring functions. By using our own scoring function (ITScore) as an example, we have shown that the method significantly improved the performance of the scoring function on binding mode and affinity predictions. Using a benchmark of 100 diverse protein-ligand complexes, the newly developed scoring function ITScore/SE yielded a success rate of 91% in identifying near-native binding modes, compared to the success rate of 82% for the original ITScore. For binding affinity prediction, ITScore/SE has yielded a correlation of R2 = 0.76 between the predicted binding scores and the experimentally measured binding affinities with the PMF validation sets of 77 diverse complexes, compared to R2 = 0.65 for ITScore. Our method can be applied to other scoring functions to account for the solvation and entropic effects.