Kinases, homology models, and high throughput docking

With the many protein sequences coming from the genome sequencing projects, it is unlikely that we will ever have an atomic resolution structure of every relevant protein. With high throughput crystallography, however, we will soon have representative structures for the vast majority of protein families. Thus the drug discovery and design process will rely heavily on protein modeling to address issues such as designing combinatorial libraries for an entire class of targets and engineering genome-wide selectivity over a target class. In this study we assess the value of high throughput docking into homology models. To do this we dock a database of random compounds seeded with known inhibitors into homology models of six different kinases. In five of the six cases the known inhibitors were found to be enriched by factors of 4-5 in the top 5% of the overall scored and ranked compounds. Furthermore, in the same five cases the known inhibitors were found to be enriched by factors of 2-3 in the top 5% of the scored and ranked known kinase inhibitors, thus showing that the homology models can pick up some of the crucial selectivity information.