We report on the prediction accuracy of ligand-based (QSAR) and structure-based (MedusaDock) methods for ranking the congeneric series of ligands binding to three protein targets (UK, ERK2, and CHK1) from CSAR. MedusaDock is designed to predict reliable docking poses ab initio. Out of the 35 cases where the unpublished co-crystal structures were resolved, 29 of our predictions (83%) were within 2.5� and 23 ligands (66%) within 2�. Therefore, MedusaDock was proven to be an efficient flexible docking approach suitable for accurate pose prediction. We have also developed predictive QSAR models of known binders of UK, ERK2, and CHK1 extracted from the publicly-available ChEMBL database. These models were used to predict binding affinity of CSAR compounds for all three targets; their accuracy was evaluated by Spearman rank correlation (0.78 for UK, 0.60 for ERK2 and 0.56 for CHK1). Finally, several consensus approaches combining MedusaDock and QSAR have been explored.

Predicting poses and binding affinity of CSAR ligands using structure based and ligand based approaches

244th National Meeting (2012)

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