Development of customized scoring function for the prediction of binding affinities of Aurora Kinase inhibitors

COMP 44

Shashidhar N Rao, rao@schrodinger.com, B. Woody Sherman, sherman@schrodinger.com, and Ramy Farid. Schrödinger, Inc, 120 W 45th Street, New York, NY 10036
We present investigations on the development of customized scoring function for the prediction of binding affinities of Aurora Kinase (AK) inhibitors, deemed to be potential anti-cancer agents. Toward this, we have docked a training set of known thiazoloquinazoline inhibitors using XP Glide protocol in the binding pocket of Aurora Kinase protein structure (2CE6). In this X-ray complex, the bound ligand (also characterized by a quinazoline moiety) demonstrates hydrogen bonding interactions with the hinge region backbone N-H, typical of many kinase inhibitors. The docked poses of AK inhibitors characterized by hydrogen bonding to hinge region N-H and hydrophobic interactions similar to the X-ray ligand were selected for two sets of calculations: (a) Prime-MMGBSA binding energy, (b) XP descriptors and (b) interaction energies with individual residues in the active site of AK. The resultant terms were employed as independent variables in the development of a scoring function to predict the pIC50 of enzyme inhibition, using multiple linear regression methods. The characteristics of the model developed and its application to a test set of AK inhibitors will be described