Doctorate of Philosophy

Abstract

Bioinformatic and epitope mapping approaches have been successful, but are reactive, in determining the mutation preferences and commonly targeted B-cell epitopes of viral fusion proteins. The primary motivation of this thesis is to describe computational methods that are proactive in predicting mutational tolerances and B-cell epitopes by assuming that the conformational rearrangements viral fusion glycoproteins undergo are one of the major fitness selection pressures that drive the evolution, especially the conservation, of viral fusion glycoproteins. The first of these methods includes the determination of mutation preferences of eight highly flexible proteins by either RECON multi-state design or single-state design using a set of discrete conformations of each protein to estimate the local physicochemical changes needed to assume multiple, low-energy conformations. This approach focused on two topics --- first, the similarity between the designed mutation preferences and natural homologs' sequence diversity, and second, the relationship of sequence conservation and stability with different aspects of protein flexibility. To address the latter topic, a new conformational dissimilarity metric was introduced, termed contact proximity deviation, which quantifies the relative changes in neighboring contacts of each residue experienced within an ensemble. The second computational method discusses how this new contact proximity deviation metric in conjunction with a residue's relative free energy score might be used as predictors of conformational B-cell epitopes, given that sites of vulnerability often overlap with sites of local conformational change that occur during viral fusion.