Computer-aided drug design has been a rapidly developing discipline that garnered increasing amount of attention in recent years. Much of these generous efforts were devoted to virtual screening of drug-size compounds. A fragment-based approach to the subject is relatively unexplored. Taking advantage of the vast amount of structural information made available by the protein databank, we designed a protocol that elucidates fragment-binding hotspots with Instance-based learning techniques. The predictions were based upon statistical inference about local chemical environment of a target protein. By analysing a thoroughly sampled conformational space of Human CDK2, we were able to arrive at a natural definition of fragment-binding hotspots, which is fragment-binding spots attested...[ Read more ]

Computer-aided drug design has been a rapidly developing discipline that garnered increasing amount of attention in recent years. Much of these generous efforts were devoted to virtual screening of drug-size compounds. A fragment-based approach to the subject is relatively unexplored. Taking advantage of the vast amount of structural information made available by the protein databank, we designed a protocol that elucidates fragment-binding hotspots with Instance-based learning techniques. The predictions were based upon statistical inference about local chemical environment of a target protein. By analysing a thoroughly sampled conformational space of Human CDK2, we were able to arrive at a natural definition of fragment-binding hotspots, which is fragment-binding spots attested against conformational change. The identified fragment-binding hotspots were validated against crystallised counterparts with consistency. Besides reporting pockets and fragments native to the existing ligand-bound CDK2 crystals, preliminary results also suggest for putative allosteric sites alongside fragments unprecedented in CDK2 crystals.