Metalloenzyme binding affinity prediction with VM2

Project summary: It is estimated that 40 to 50% of known enzymes can be characterized as metalloenzymes, while currently only 7% of FDA-approved drugs in the United States target this class of protein. This is despite the fact that there are many dozens of already identified metalloenzyme targets involved in virtually every therapeutic area, including anti-inflammatory, antibiotics, antivirals, anticancer drugs, and more. This is in large part because the already very difficult drug design requirement to maintain/increase the potency of an initial ligand (drug-like molecule) while improving/maintaining its target selectivity and pharmacokinetic properties, is made even harder by the complicated and often non-intuitive nature of metal-ligand and metal-protein interactions. Accurate molecular modeling predictions of metalloenzyme-ligand binding affinities, then, would be highly impactful in pharmaceutical industry drug research and development programs, because they would allow R&D scientists to carry out computational experiments drastically reducing the number of expensive and time-consuming bench experiments required to overcome the difficult metalloenzyme inhibitor design challenges they face. However, currently available molecular modeling approaches are unable to make predictions reliable enough to do this. Docking and scoring methods are able to determine, in many cases, the pose of inhibitors in metalloenzyme active sites, but they cannot correctly rank candidate inhibitors in order of binding affinity as they lack the required detail in their energy models. Recently, free energy-based methods have advanced to the point of providing reliable binding affinity predictions for many non-metal protein-ligand series and can, therefore, help speed ligand discovery efforts for these systems. They cannot provide good binding affinities for metalloenzyme-ligand systems though, because to-date they are all entirely based on classical forcefields, which fundamentally limits the accuracy of their descriptions of metal-ligand and metal-protein interactions. This is due, in part, to lack of inclusion of important polarization and charge transfer effects, but it is also because the complex electronic structure, which metals often exhibit, is intrinsically quantum mechanical. This fast-track SBIR proposal will address this by developing a new and unique molecular modeling software tool called Mzyme-QM-VM2, which will provide reliably accurate binding free energies for metalloenzyme- inhibitor complexes by a novel combination of statistical mechanics and highly scalable quantum chemistry methods. This software will be based on mining minima free energy calculation methodology and will be developed as an extension of VeraChem's VM2 free energy software platform.