"A QSAR Model for Prediction of Kinase Pharmacological Similarity Based on Profiling, Sequence, and 3D Structural Data "

Wendy Cornell
Merck Research Laboratories, USA

We propose a direct QSAR methodology to predict how similar the inhibitor-binding profiles of two protein kinases are likely to be based on the similarity of the properties of the residues surrounding the ATP-binding site. A model is produced for two sets of data: Kd's published by Karaman et al (Nature Biotechnology 2008, 26, 127-132) for 38 compounds on 200 kinases, and percent inhibition data on up to 900 in-house compounds tested on 30 to 51 kinases by the Dundee University DSTT Consortium. Each model is self-consistent by cross-validation and both models point to only a few residues in the active site of kinase controlling the binding profiles, however they do not agree on which residue is most important. We apply each model to predict the similarity in binding profile to all pairs in a set of 411 kinases from the human genome. While we do not believe either model is definitive, the approach is promising and can be applied to larger and better datasets when they become available.

<<voltar