Scoring function is a popular computational method for predicting the binding affinity of a protein-ligand complex. It usually combines with molecular docking and has very wide applications in structure-based drug design. However, there are still several key factors which hinder the further improvement on the prediction accuracy of current scoring functions, i.e. the desolvation energy, etc. Our research aims to develop a three step strategy for computing the desolvation energy by using the information of predicted explicit water molecules. First, the PMF parameters are derived for describing the protein-water interactions, on the basis of a non-redundant data set of crystal structures with water molecules. Then with these parameters, a largest-neighbored clustering algorithm is used to predict the potential water sites in the protein binding pocket. Finally, the desolvation energy is obtained with the "displaced-solvent" method. We integrate the desolvation energy with the scoring function X-Score and optimize the customized weight factors of each energy term via a Monte Carlo Simulated Annealing (MCSA) algorithm. The new scoring function demonstrates improved performance on binding affinity prediction on several data sets of both specific and diverse targets.

Improving the scoring of protein-ligand binding affinity by including the desolvation energy

251st National Meeting (2016)

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