Protein kinases: docking and homology modeling reliability.

A database of about 700 high-resolution kinase structures was used to test the reliability of 17 docking procedures (using six docking software packages) by means of self- and cross-docking studies. The analysis of about 80 000 docking calculations suggests that the docking of an unknown ligand into a kinase has a probability of only 30-37% to be a correct ligand pose. However, based on the hypothesis that docking calculations are more reliable if the ligand to be docked is similar to the ligand present in the complex from which the target docking protein has been extracted, we propose an automated procedure that is able to improve the docking accuracy, suggest the best protein for docking studies, and assess the statistical reliability of docking calculations. The results were also transferred to the homology modeling field and led us to propose an alternative strategy based on ligand similarity for the development of kinase models whose experimental structure was not known. Our results suggest that in many cases this approach can give better results than the classical homology modeling procedure based exclusively on the sequence homology.