| A computational geometry technique employing Delaunay tessellation of protein structure represented by Calpha atoms to derive a statistical residue contact potential is used to analyze the structure-function relationship in beta-strand proteins.; In Chapter 2, we applied the computational geometrical approach to identify geometric, invariant core of the protein family, that is, the invariant characteristics that are shared by all members of the family. Independent of any coordinate system, our core-finding method does not require either the procedure of superposition or the prior assignment of secondary structure for the residues of the protein.; In Chapter 3, we used the computational geometry approach to align the proteins structurally. Alignment was applied on proteins with similar structures but with sequence identities varying from >10% to <70%. This approach essentially reduces the problem of structural alignment to the problem of sequence alignment, alignment of the residual scores of multiple proteins. This simple method avoids much of the computational complexity of other methods discussed in detail in the chapter.; In Chapter 4, we studied the effects of single residue mutations on the enzymatic activity of kinases B-raf, BCR_ABL and Fyn tyrosine kinase. Structural changes due to amino acids substitution were quantified using total potential score, and the resulted total score of the proteins were correlated with the kinase activity determined in experiments and obtained from published results. This simple method was able to predict the kinase activity after specific point mutations.; Chapter 5 describes the machine learning models we developed to characterize the enzyme mutant affect. Models achieve mean accuracy of 80.2% with support vector machine and 85.9% with random forests. Training set properties are examined using randomized controls and learning curves. This work improves model performance and predictive capability by including additional components to the mutant residual profiles that incorporate additional sequence and structure information.; In Chapter 6, we used the computational geometry approach to study the strands stability of Ig27 protein during molecular dynamics (MD) simulation. It is known experimentally (using the atomic force microscopy (AFM) for example) that the Ig27 protein unfolding pathway is as follow: A → A' → G → F → B → C → D,E. However, the relevance of these experiments to the behavior of the protein under physiological conditions was not established. (Abstract shortened by UMI.)... |
