セミナー2017 –

Protein kinases are highly perspective drug targets since they are implicated in critical functions in signalling pathways in all cells. In many cases development of highly-specific kinase inhibiting reagents is crucial [1]. Nevertheless, development of drugs based on kinase inhibitors is hindered by the problem of selectivity causing numerous off-target effects. Kinase proteins binding pockets are highly conservative structurally which makes developing of the inhibitors selective to a particular target against other kinases a highly challenging task. Based on structural kinase dataset (109 kinase structures with bound ligands) we have performed analysis of structural pocket similarity and sequence similarity and analyzed the association of these pairwise similarities with activity correlations (activity data from the PubChem BioAssay Database dataset ‘Navigating the Kinome’ [2]). We have found that for many kinases this association is very strong and can indicate close related proteins. We have developed a method for kinase prioritizing and activity prediction for a novel drug-like compounds for purposes of design of small-molecule highly-selective kinase inhibitors. In our approach, we apply machine learning techniques on the interaction potential descriptors for protein-ligand complex structures obtained by docking and calculate several descriptors of different types to characterize protein-ligand interactions for a given ligand conformation. We demonstrated that our approach is unbiased to the ligand structural similarity to compounds in the training set, which is a common issue in classical QSAR methods. We believe our approach to be suitable for a high-quality prediction for a novel leads in an application for kinase-targeted drug design. (Ref. [1]Nature Reviews Cancer, 9(1), 28–39 (2009)., [2] Nature chemical biology, 7(4), 200-202 (2011).)

The controlled activation of proteins in living cells is an important goal in protein-design research, but to introduce an artificial activation switch into membrane proteins through rational design is a significant challenge because of the structural and functional complexity of such proteins. Here we report the allosteric activation of two types of membrane-bound neurotransmitter receptors, the ion-channel type and the G-protein-coupled glutamate receptors, using coordination chemistry in living cells. The high programmability of coordination chemistry enabled two His mutations, which act as an artificial allosteric site, to be semirationally incorporated in the vicinity of the ligand-binding pockets. Binding of Pd(2,2′-bipyridine) at the allosteric site enabled the active conformations of the glutamate receptors to be stabilized. Using this approach, we were able to activate selectively a mutant glutamate receptor in live neurons, which initiated a subsequent signal-transduction pathway.(Ref. Nat. Chem., 8, 958–967 (2016).)