| The general goal of this work was to improve the reliability of both homology modeling of proteins and docking of ligands to homology models. The main factors reducing the accuracy of current homology modeling methods are errors in alignment and poor prediction of regions structurally dissimilar to template. In the first part of this dissertation, methods to address both these problems are presented. An algorithm, Consensus, was developed that provides highly accurate target-template alignment and identifies regions of poor alignment reliability and potential structural dissimilarity. The method was implemented as the "Consensus" server. The framework regions thus identified are modeled highly accurately. Since no methodology can eliminate all potential errors, use of multiple models is proposed. Once a framework model is generated using the Consensus method, the remaining regions are modeled by exploring a variety of alignments and conformations. The resulting ensemble of models clearly shows the uncertainty of the modeling in regions without sufficient structural similarity to the template.; In the second part of this work, an approach of multiple docking to multiple homology models was explored. Retaining large sets of docked conformations reduces the reliance on scoring functions and shows the distribution of most likely ligand positions, thereby improving the robustness of results. This approach was applied to study the substrate specificity differences between Cytochrome P450 1As (CYP1As) in fish and humans. Three carcinogenic compounds 3,3',4,4'-tetrachlorobiphenyl (TCB), benzo-[a]-pyrene (BAP) and 2,3,7,8-tetrachloro-dibenzo-p-dioxin (TCDD) are known to be metabolized at vastly different rates and with different regiospecificities by fish and human CYP1As. In order to understand the structural basis of these phenomena, the three molecules were docked to large sets of homology models of the enzymes. The resulting distributions of docked conformations of TCB explained the differences in the rates of its metabolism observed between the species. Further analysis identified the putative amino acid residues responsible for these binding differences. The BAP orientations among the productive conformations agreed with the experimentally observed metabolites. The differences in metabolism for TCDD between the species have also been predicted based on docking results. |
