Computational techniques for modeling protein -ligand interactions and their application to serine proteases and asparaginyl -tRNA synthetase

This thesis describes several techniques for modeling protein-ligand interactions, including interactions between water molecules and proteins, as well as application of these techniques to thrombin, trypsin, and asparaginyl-tRNA synthetase. A complete-linkage hierarchical cluster analysis technique for determining the degree of conservation of water sites in a series of related protein structures is presented. This technique was applied to the study of conserved water binding sites in the serine proteases thrombin and trypsin, with analysis of the implications for conserved water sites in the active site and nearby sodium ion site in thrombin. Cluster analysis was also used to identify conserved water binding sites in bovine pancreatic trypsin inhibitor (BPTI) and in the trypsin-BPTI complex. The conserved sites in the trypsin-BPTI complex were compared to those in trypsin and BPTI, showing that only about half of the interfacial water sites in the complex exist in the free form of either protein, while the remaining half are recruited or shuffled upon complex formation. The results of cluster analysis also allow inclusion of highly conserved water sites in protein and drug design.;In addition to examination of water molecules as protein ligands, modeling sites of favorable potential interactions with proteins in the SLIDE technique for computational ligand screening was improved. S LIDE is a multi-step algorithm which eliminates potential ligand molecules from a screening database via increasingly more stringent and computationally expensive steps to yield a ranked list of potential ligands for the protein target. Improvements made to the modeling of sites of favorable hydrophobic interaction for both the protein template and the potential ligand molecules are described and evaluated. Additional improvements made by my colleague Maria Zavodszky to the description of hydrogen bonding points in the protein template are also briefly described and evaluated. These improvements were tested using 42 thrombin and 16 glutathione S-transferase complexes. Both the protein template and database molecule representation improvements yield ligand dockings that are closer to those seen in the crystallographic complex. An enrichment of known ligands selected from a set of molecules from the Cambridge Structural Database (CSD) is also shown.;Application of SLIDE to asparaginyl-tRNA synthetase from Brugia malayi, a human pathogenic nematode, was performed. Screening against the CSD identified three potential ligands of particular interest: variolin B, with possible antitumor and antiviral properties; cercosporamide, which has known phytotoxic and fungitoxic properties; and phlorizin, a sodium/glucose transport inhibitor. Suggested binding modes for two of 16 in vitro high-throughput screening hits are also described.