The overall aim of this project is to develop a novel integrated suite of software applications for predicting, refining and manipulating biomacromolecular structure, particularly with respect to computational virtual screening of drug candidates. Virtual screening of a panel of ligands against a biomacromolecule target requires a highly accurate model for the ligand binding site, as well as rapid and effective estimation of binding affinity. In many important drug discovery projects both of these requirements cannot be met. The two main goals of the proposed research are thus to: 1) improve the accuracy and relability of free energy scoring of putative protein-ligand complexes, and 2) enhance the quality of low-resolution structural models from x-ray, nmr or comparative modeling to make them useful as targets for virtual screening. The core technology for this research plan is based on the HINT (Hydropathic INTeractions) program/paradigm that exports (with reasonable speed and accuracy) both a unique empirical free energy forcefield and threedimensional graphics objects that encode significant structural information. A number of specific software tools will be created by this effort: a) an integrated docking system using HINT forcefield scoring;b) an automated computational titration program that evaluates the ionization state of residues and ligand functional groups to optimize ligand binding;c) a range of methods to predict and/or optimize water molecule locations in environments where water-mediated hydrogen bonding could impact ligand binding;d) a new de novo ligand design protocol based on three-dimensional hydropathy maps;and e) integrated crystallographic and NMR refinement program(s) using the hydropathic forcefield as a target function. This latter tool may be extended as an adjunct to homology modeling approaches to creating target structures, and may prove useful for defining models of inaccessible proteins. Specific collaborative arrangements are in place to apply these tools to a range of current drug discovery problems.
Showing the most recent 10 out of 23 publications
