Automating and improving virtual screening of large compound databases


Niu Huang1, John Irwin2, Chakrapani Kalyanaraman2, Brian Shoichet2, and Matthew P Jacobson2. (1) Department of Biopharmaceutical Sciences, University of California, San Francisco, 600 16th St., Suite N474E, San Francisco, CA 94143, (2) Department of Pharmaceutical Chemistry, University of California, San Francisco, 1700 4th St, San Francisco, CA 94143
Despite well-known weaknesses, molecular docking is now one of the most practical techniques to leverage structure for ligand discovery. We have developed a docking and rescoring protocol to computationally screen chemical databases containing millions of compounds with minimal user intervention. Docking was performed using the DOCK 3.5.54 program with a grid-based electrostatic and van der Waals interaction energy evaluation including a partial ligand desovlation energy correction. This fully automated docking approach was evaluated by the extent to which known binders were enriched against a background of drug-like decoys and compared favorably with enrichments obtained by an expert. The binding poses of top scoring compounds from docking was submitted to further refinement and rescoring using an all-atom force field (OPLS AA) and implicit solvent model (Generalized Born); the use of a rapid multi-scale Truncated Newton energy minimization algorithm enabled this refinement stage to be completed with less than one minute per ligand. Significant improvement in enrichment was observed for most of the systems studied.

Advances in Virtual High-Throughput Screening
8:20 AM-11:30 AM, Wednesday, August 25, 2004 Pennsylvania Convention Center -- 110A&B, Oral

8:00 PM-10:00 PM, Monday, August 23, 2004 Pennsylvania Convention Center -- Hall D, Sci-Mix

Division of Chemical Information

The 228th ACS National Meeting, in Philadelphia, PA, August 22-26, 2004