Computational Approaches for Mutation Phenotype Prediction and Protein Binding Site Characterization

Many genetic diseases result from local variations in the sequence and structure of proteins that play essential roles in the human body. Therefore, it is relevant to ask how local variations in sequence and structure influence the molecular function of a protein. This thesis consists of two projects, both of which facilitate our understanding of the sequence-structure-function relationship in disease-related proteins.;Single amino acid mutations are often tolerable, but some lead to drastic disruption of protein structure or function. For some genetic diseases, effective treatment depends on early detection of disease-causing mutations. Prediction of the disease or neutral phenotypes of single amino acid mutations should be based on knowledge of protein structure, function, and the relationship between the two. Chapter 2 describes a novel approach for mutation phenotype prediction, which aims at improving prediction accuracy in phenylketonuria, based on the hypothesis that a prediction model trained on a single protein family would perform better on prediction tasks in the same protein.;Ligand binding is the most fundamental mechanism by which proteins carry out their functions. Therefore, computational characterization of ligand binding sites provides information for relating function to structure. This motivates the development of methods and tools for advanced characterization of ligand binding sites. Chapter 3 describes SiteComp, a web server which provides three types of analyses on known binding sites: (i) structural comparison of two similar sites to identify and highlight differences in molecular interaction properties; (ii) evaluation of the contribution of individual amino acid sidechains to protein-ligand interaction; (iii) identification of sub-sites with distinct molecular interaction properties within a larger binding site. These analyses are compared with exiting methods for similar or related purposes, and their uniqueness and advantage are discussed in detail. As part of the development of SiteComp, a previously developed server for binding site identification was updated to enable seamless analysis of binding sites from identification to advanced characterization (Appendix C).;Taken together, the approaches and tools presented here help advance our understanding of the sequence-structure-function relationship in disease-related proteins.