Scoring functions for binding affinity prediction based on the structure of a protein-ligand complex have had some success, but suffer from inadequate treatment of key aspects of the physics of binding. Three major areas where improvement is needed is in the treatment of water structure, ligand strain energy, and protein reorganization energy. Based on an analysis of a large data set of PDB structures containing targets for which multiple, diverse ligand cocrystals are available, we have developed an improved version of our Glide XP scoring function which provides significantly better results for both rank ordering of active compounds and discrimination of active compounds from a random database of drug-like molecules. Results for ~30 targets for both of these capabilities will be discussed. We will also discuss the use of our WaterMap molecular dynamics simulation technology to specify the water structure in the active site, and diverse applications of this methodology to a wide range of ligand series.
Back to Workshop I: Design of Drugs and Chemicals that Influence Biology