We have been interested in using molecular mechanics force fields and implicit solvent models as scoring functions for protein-ligand docking, a strategy that is intermediate in computational expense between typical empirical scoring functions and rigorous free energy calculations.  In general the approach works best for ranking chemically similar compounds, such as those in congeneric series.  In the most recent prediction exercise from Community Structure-Activity Resources (CSAR), we applied this approach to several series of ligands.  In each case, we used a crystal structure with one member of the series bound to predict the conformations and relative affinities of the other members of the series.  Our success in ranking compounds according to potency varied among the data sets.  For LpxC, we found it very easy to distinguish actives from inactives based on our predicted docking scores.  For urokinase, ERK2, and Chk1 kinase, our predicted scores showed some correlation with experimental potency, generally with r2 ~ 0.3.  Importantly, the compounds with the most favorable scores were generally among the most potent.  Outliers will be analyzed with the goal of improving the predictive methods.

Predicting relative binding affinities of congeneric series of ligands: A prospective docking study using Prime docking

244th National Meeting (2012)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.