Targeting the inactive conformation of protein kinases: computational screening based on ligand conformation

Recent strategies to develop potent and selective inhibitors of protein kinases have targeted a pocket exposed by many kinases in their inactive conformation. In order to identify these Type H kinase inhibitors, we have developed a computational screen using combined pharmacophore- and shape-based criteria. Here, we successfully discriminate Type II kinase inhibitors from a large number of Type I and other decoy ligands taken from the MOE kinase and DUD databases. The pharmacophore component of the screen is effective in limiting false positives, whilst ROCS rapidly prioritizes Type II kinase inhibitors for this analysis. Here we also show that the screen does not appear strongly biased towards particular kinase target types, and can detect Type Ifs of diverse linker chemistry. Analysis of this large scale validation not only identifies known Type II kinase inhibitors but additionally identifies DUD compounds with relatively low scaffold similarity to known Type II kinase inhibitors but high similarity in 3D conformation. Our method therefore uses 3D structure to detect putative active compounds which would not have been detected by a 2D-only approach. The method therefore shows promise as a tool in the discovery of inhibitors with potentially novel scaffolds that target protein kinases in their inactive conformational state.