Q-MOL: High Fidelity Platform for In Silico Drug Discovery and Design
For several decades, in silico drug discovery methods have held the promise of cheap and efficient ways of streamlining de novo discovery of small molecule ligands against novel drug targets. The existing computational methodologies proved to be successful in those cases when a protein target is a relatively rigid molecule (e.g., many enzymes), or when an in silico methodology is applied during hit-to-lead optimization of an already potent binder. However, these methods fail when applied to highly flexible or intrinsically disordered protein targets which represent the bulk of pharmacologically important proteins. The major factors contributing to the failure are either incorrect or simplified treatment of protein flexibility, inadequate or overfitted scoring function, absence of correctly annotated druggable pockets, and lack of structural data about potential allosteric sites. The Q-MOL drug discovery platform addresses the problem of treatment of protein flexibility during the course of protein-ligand docking by computationally implementing some of postulates of the energy landscape theory of protein folding. The Q-MOL protein-ligand docking method was thoroughly validated in academic and commercial settings across a panel of more than 60 diverse proteins (of viral, bacterial and animal origin). This methodology, originally developed for high fidelity virtual ligand screening, is also used for the prediction of ligand binding sites. The ability to reliably predict ligand binding sites on the surface of novel and unannotated protein targets is critical for successful drug discovery. The Q-MOL drug discovery platform is agnostic to the nature of a protein target. Rigid, flexible or intrinsically disordered proteins are processed by the same protein-ligand docking protocol. Importantly, the software does not require the presence of structurally well-defined binding pockets or sites. The correct and validated implicit treatment of protein flexibility allows for successful discovery of small molecule ligands against allosteric sites lacking any structural features of a classical druggable pocket. To demonstrate the power of its methodology, Q-MOL was used to detect ligand binding sites on the surface of the bHLHZ domain of the c-Myc protein. Q-MOL virtual ligand screening was then used to dock a human metabolites library against one of the identified binding sites. The top hits were cross-referenced against published data, notably identifying dehydroascorbic acid as one of the top-ranked compounds. In contrast to other docking programs, all of the Q-MOL in silico drug discovery features were thoroughly validated in vitro, in cellulo and in vivo when possible. In this paper, the results of several previously published drug discovery projects are briefly presented and related to relevant computational features of the software. The corresponding protein targets include viral proteinases (West Nile, Hepatitis C, Dengue and Zika viruses), hemopexin domain of membrane type I matrix metalloproteinase, retinoid X receptor , and {beta}-catenin armadillo repeat domain. It has been also shown that contrary to existing belief, highly flexible proteins (or their domains) represent significantly more amenable drug discovery targets than rigid proteins. Flexible and intrinsically disordered proteins exist in a vast array of environment-dependent conformational states, and thus, they can accommodate a significantly larger number of ligand chemotypes ensuring both potency and specificity of interactions with a protein target. It has been also demonstrated that Q-MOL binding sites prediction and ligand docking methodologies, initially developed for protein targets, can be successfully applied "As Is" to polynucleotide-based structures, such as non-coding viral RNA molecules.