| Katarzyna Bernacki1, Chakrapani Kalyanaraman1, Ilya Chorny2, and Matthew P Jacobson1. (1) Department of Pharmaceutical Chemistry, University of California San Francisco, Box 2240, San Francisco, CA 94143-2240, (2) Department of Pharmaceutical Chemistry, University of California, San Francisco, San Francisco, CA 94143-2240 |
| Protein homology models have been used in conjunction with virtual screening to successfully identify novel inhibitors over the past few years. However, it is widely accepted that docking to homology models is more challenging and less successful than docking to crystallographic structures. We are developing methods to improve the quality of homology models for in silico screening, especially by energy-based refinement of binding site residues. The homology modeling and refinement protocols use an all-atom force field (OPLS-AA) and a Generalized Born implicit solvent model; docking is performed with Glide. Test cases so far include DHFR, CDK2, several parasitic cysteine proteases, human lipoxygenase, and alphaVbeta6 integrin. In all cases, we demonstrate that the enrichment of known inhibitors can be improved by refinement of the binding site, and that the refined homology models (at ~50% sequence identity) perform comparably to crystal structures. In two cases so far (alphaVbeta6 integrin and human lipoxygenase), experimental collaborators have identified new inhibitors based on our docking results, with hit rates comparable to docking against crystal structures. |
