Algorithm development for flexible molecule docking: Application to drug design and macromolecular complex prediction

Data produced from structural biology projects afford us opportunities to understand the relationship between molecular structure and function to a greater extent than ever before. Despite the power of today's computers, the challenge of modeling molecular interactions presents a problem of such complexity that computers hundreds of times faster would still not be able to completely sample all possible states of interacting molecules. Here, I explore new algorithms for the inclusion of ligand flexibility and the limited combinatorial elaboration of chemical space. The feasibility of using pre generated conformational ensembles of ligands as proxies for docking flexible ligands is explored. The ensemble method is tested against a panel of well-studied model systems and is observed to increase the calculation speed by up to 100-fold. In collaboration with a pharmaceutical company, the new algorithm is used to discover new drug leads for a clinically relevant target. In comparison to the industry-standard, high-throughput screening, the docking calculation outperforms the experimental screen by more than 180-fold.; Molecular docking calculations scale as the power of the number of rotatable bonds in the system. Systems with a dozen rotatable bonds could easily take millions of years to calculate. A combinatorial method is introduced for docking flexible molecules that flattens this exponential time dependence and makes feasible calculations that otherwise would have taken more time than the universe is thought to have existed. Additionally, the combinatorial method extends the purview of the docking algorithm into the domain of chemical space exploration. The combinatorial method is applied to the protein-protein docking problem, predicting the juxtaposition of biological macromolecules, and evaluating point mutations on proteins. An appendix to the work examines the effect of clustering chemical libraries to increase diversity and identifies novel inhibitors of well-studied drug targets.