Novel docking and scoring approach to rank-ordering compounds by potency

COMP 152

Xuan Hong, Xuan.2.Hong@gsk.com1, Scott Kahn, scott@accelrys.com2, C. M. Venkatachalam, venkat@accelrys.com1, Chaya Duraiswami, Chaya.2.Duraiswami@gsk.com3, Catherine E. Peishoff4, and Martha S. Head4. (1) Accelrys, Inc, San Diego, CA 92121, (2) Accelrys Inc, 9685 Scranton Road, San Diego, CA 92121, (3) Computational and Structural Sciences, GlaxoSmithKline Pharmaceuticals, 1250 South Collegeville Road, UP-1110, Collegeville, PA 19426, (4) Computational, Analytical and Structural Sciences, GlaxoSmithKline Pharmaceuticals, Research and Development Division, P.O. Box 5089, Collegeville, PA 19426-0898
Docking calculations are typically used for three purposes: (1) predicting the binding modes of cognate ligands in protein structures, (2) high throughput docking to identify active compounds for a protein target, and (3) rank ordering a set of closely related analogs by their potency. While many docking and scoring programs are able to predict binding modes of cognate ligands with satisfying correctness and can be successfully used for lead identification, to the best of our knowledge very few programs can rank order analogs by their potency with good accuracy. We will present a novel approach aimed at using the results of a docking calculation to estimate binding affinities. It seeks to incorporate contributions to binding free energy that are not adequately described by typical scoring functions. These contributions include the consideration of (1) multiple ligand conformations to binding free energy, (2) solvation and desolvation processes taking place upon binding, and (3) entropy change upon binding. Several novel descriptors were developed to characterize these contributions. We will describe our modeling strategy and its application to specific test cases. Results will also be presented with a comparison of rank-ordering of the compounds using our scoring function against their potency.