Ensemble methods in computational protein and ligand design : applications to the Fc[gamma] immunoglobulin, HIV-1 protease, and ketol-acid reductoisomerase systems

Abstract

This thesis explores the use of ensemble, free energy models in the study and design of molecular, biochemical systems. We use physics based computational models to analyze the molecular basis of binding affinity in the context of protein-protein and protein-ligand binding as well as reaction rate enhancement in enzyme catalysis. First, we evaluate the solvent screened energetics of immunoglobulin G (IgG):Fc[gamma] receptor binding using molecular mechanics, Poisson-Boltzmann surface area (MMPBSA) models. We assess the role IgG1 linked glycans play in binding to human Fc[gamma]-III and computationally evaluate experimentally designed Fe mutations that recover binding affinity in the absence of glycosylation. Using the insight gained from this study, we developed novel murine IgG variants with engineered Fc[gamma] receptor binding patterns via the computational design and experimental validation of Fc mutations that are predicted to knock out binding to Fc[gamma]R-IV. Our design and analysis highlight the importance of solvent screened electrostatic interactions and electrostatic complementarity in protein-protein binding. Second, we develop novel, ensemble methods to measure configurational free energy and entropy changes in protein-ligand binding and use it to predict the relative binding affinity of a series of previously designed HIV-1 protease inhibitors. We find that using configurational free energies to evaluate inhibitor efficacy significantly improves relative ranking of inhibitors over traditional, single-point energy metrics, but that only a relatively small number of low energy configurations are necessary to capture the ensemble effect. Finally, we present a joint study of the redesign and dynamic analysis of ketol-acid isomeroreductase (KARI). We first develop and apply a novel, end-point method to rationally design enzyme variants that reduce the free energy of activation, and present the computational and experimental analysis of a series of designed KARI mutants. Our analysis reveals that this transition-state theory based approach is effective at reducing the enthalpy of activation, but also increases entropic activation penalties that ultimately overpower the enthalpic gains. A dynamic analysis of these KARI variants is also presented, in which the transition path ensemble is explored using transition path sampling. We find that this ensemble approach is better able to predict relative enzyme activities and suggests a conserved, dynamic mechanism for catalysis. The results and analysis presented herein demonstrate novel, computational approaches to account for ensemble effects in the study and design of effective biomolecules.