Computational ligand design and analysis in protein complexes using inverse methods, combinatorial search, and accurate solvation modeling

Abstract

This thesis presents the development and application of several computational techniques to aid in the design and analysis of small molecules and peptides that bind to protein targets. First, an inverse small-molecule design algorithm is presented that can explore the space of ligands compatible with binding to a target protein using fast combinatorial search methods. The inverse design method was applied to design inhibitors of HIV-1 protease that should be less likely to induce resistance mutations because they fit inside a consensus substrate envelope. Fifteen designed inhibitors were chemically synthesized, and four of the tightest binding compounds to the wild-type protease exhibited broad specificity against a panel of drug resistance mutant proteases in experimental tests. Inverse protein design methods and charge optimization were also applied to improve the binding affinity of a substrate peptide for an inactivated mutant of HIV-1 protease, in an effort to learn more about the thermodynamics and mechanisms of peptide binding. A single mutant peptide calculated to have improved binding electrostatics exhibited greater than 10-fold improved affinity experimentally.

(cont.) The second half of this thesis presents an accurate method for evaluating the electrostatic component of solvation and binding in molecular systems, based on curved boundary-element method solutions of the linearized Poisson-Boltzmann equation. Using the presented FFTSVD matrix compression algorithm and other techniques, a full linearized Poisson-Boltzmann equation solver is described that is capable of solving multi-region problems in molecular continuum electrostatics to high precision.