Improved optimization of covalent ligands using a novel implementation of quantum mechanics suitable for large ligand/protein systems.

Project Summary The value of computational chemistry to commercial drug discovery is now well-established. Virtual screening (including molecular docking) now jumpstarts most discovery efforts. Tools such as molecular dynamics and free energy perturbation are increasingly used to inform the later stages of lead refinement. The growing importance of computational structure-based methods has influenced the types of ligands that are identified. The energy of a molecular system is fully described by quantum mechanics (QM). However, QM equations are extraordinarily complex, and applying QM to realistic models of relevance to drug discovery on a suitable timescale has traditionally been impossible. Instead, a simplified formulation of molecular interaction, molecular mechanics (MM), has been used. The analytic equations of MM can be easily assessed directly from the coordinates of a molecular structure. However, MM suffers severe limitations relative to the QM representation, including poor estimation of certain types of molecular effects (polarization, π-stacking, and interactions with metals and halogens) and an inability to deal with changes in topology, including bond creation/breakage. Because of this latter limitation, drug discovery in the computational era has focused largely on non-covalent inhibitors. However, covalent drugs are historically significant (aspirin, penicillin, more than 50 FDA approved drugs in total). A growing realization that covalent drugs can provide a way to address problems that non- covalent ligands cannot address has led to a resurgence in interest in drug covalency. Among the targets that are especially well suited for covalent drugs are: drugs that differentiate among similar binding sites (e.g., the Kinase family); Protec drugs that can lead to protein degradation; and ligands that can target “undruggable” targets such as protein-protein interactions. In turn, this realization has led to renewed interest in QM methods. We recently described a new, novel implementation of QM that (for the first time) allows accurate DFT/QM to be applied to large ligand/protein systems with sufficient throughput for drug discovery. This new approach allows calculations to be carried out in less than an hour on a massively distributed computing platform, as compared to weeks or months using traditional QM implementations. This makes it possible to use QM-based computational tools to optimize covalent ligands--including such previously elusive goals as tuning the “warhead” reactive group on the ligand. Subsequent work we have carried out has further demonstrated the ability of QM to improve upon standard scoring approaches for covalently-bound ligands. This has led us to develop an approach that will streamline and optimize the process of computationally-driven covalent ligand characterization. The result will be a QM approach that can reliably focus ligand optimization—including the warhead—on a timescale commensurate with modern drug discovery.