Novel Consensus Docking Strategy to Improve Ligand Pose Prediction

Abstract

Molecular docking, which mainly includes pose prediction and binding affinity calculation, has become an important tool for assisting structure-based drug design. Correctly predicting the ligand binding pose to a protein target enables the estimation of binding free energy using various tools. Previous studies have shown that the consensus method can be used to improve the docking performance with respect to compound scoring and pose prediction. In this report, a novel consensus docking strategy was proposed, which uses a dynamic benchmark data set selection to determine the best program combinations to improve the docking success rate. Using the complexes from PDBbind as a benchmark data set, a 4.9% enhancement in success rate was achieved compared with the best program.