Theory and Modeling of Noncovalent Binding

项目总结

Project Summary Identifying a small molecule that tightly binds a targeted protein is a time-consuming, costly step in many drug discovery projects. Explicit solvent free energy methods can be used to predict small molecule-protein binding affinities and thus assist with this step. However, they do not provide consistently accurate predictions, and limitations in the force fields they use are implicated as a key source of error. Our main goal, therefore, is to help generate more trustworthy force fields. In particular, we aim to prove principle for the use of experimental binding data for host-guest systems, along with traditionally used liquid properties, to refine force field parameters. We also aim show that free energy methods can help predict ligand binding poses and rank compound libraries against targeted proteins. First, we will expand the chemical diversity of host-guest systems, by developing facile methods of derivatizing cyclodextrin host molecules, and using these methods to create new, water- soluble cyclodextrin derivatives. We will measure their binding free energies and enthalpies with varied guest molecules, and will use these new data to test and refine force fields. We also aim to prove principle for the use of sensitivity analysis to refine Lennard-Jones (LJ) parameters in existing atom-typed force fields, based on host-guest binding data and liquid property data. In addition to adjusting existing atom-typed parameters, we will develop an atoms-in-molecules approach to mapping a quantum calculation for a molecule to LJ parameters for that molecule. By reducing the number of parameters, relative to atom-typed methods, this approach should enable global parameter optimization, rather than just refinement of existing parameters. Finally, we will automate and optimize our lab’s attach-pull-release (APR) method of computing binding free energies so that it can be used to rank candidate poses of a ligand in a binding site; the most stable few poses will then be used for full binding free energy calculations. Success in this effort will enable free energy methods to be used in virtual compound screening. In addition, we will use the APR method to test the new parameters generated above in the context of protein-ligand binding.