Computational study of protein interactions

Abstract

Motivation Protein molecules are the physical entities that mediate cellular functions such as duplication of genes, expression of genes, and transduction of information within and between cells. Despite of the variety of the function, the physical nature of these activity can be simply explained by the interaction of protein molecules and their ligands. For example, in cellular signaling process a protein kinase interacts with a cofactor ATP, recognizes its partner protein, and transfers a phosphate group to the partner. Therefore the understanding of the interaction between protein molecules and their ligands would lead to deeper understanding on how the biological processes take place, and that may enable us to control the biological processes for human purposes. The interaction of protein molecules with their ligands can be studied with different methodologies. The binding affinity of protein molecules and their ligand molecules can be experimentally measured. The conversion rate of a certain chemical compound can be measured to study the functionality and mechanism of an enzyme. However, the three dimensional structure of the protein molecule in complex with its ligand provides most ample and complete information on the mechanism of the interaction. Moreover the structural study of protein interaction may provide hints that may enable us to modulate the function of proteins. In most cases, the modulation of the functionality requires a structural model for proteins that will enables a rational guesses for the modulation. Generally these structural models were developed by referring the existing structures of the protein ？ ligand complexes. These models can be utilized to achieve practical goals such as developing new medicines and new enzymes. In this thesis, the interaction of the protein has been computationally analyzed and it has been modeled with different structure models. As the results, the structures of proteins were studied in three different aspects: the interaction with water molecules, the effect of structural ensemble in binding affinity calculation, and modeling of the protein surface region to find a highly variable and accessible surface.

Results In the first study, I focused on the local protein structures that interact with water molecules. Accordingly, the water molecules observed in the X-ray crystallography and nearby protein structures around these water molecules were subjected to the analysis. Especially the hydrogen bonding structure around the water molecules had been analyzed to found out structural features of water interacting local structure, which can further be used in designing of protein-protein and protein-ligand interaction structure. This analysis provided evidences that the hydrogen bond structure of the protein associated water molecules might be different from the tetrahedral geometry, which have been thought to be the hydrogen bond structures around water molecules due to its energetic stability. Rather, planar structures were observed along with preferentially hydrogen bonding with hydrogen acceptors while balance between donor and acceptor is expected in the tetrahedral geometry. This structure would be more similar to the structure of liquid water where the rotational entropy would have significant contribution. Additionally the preference of the water molecules to the protein-protein interface was observed, and the preference toward complex interaction structures was found in the biological protein-protein interfaces. A new water mediate interaction potential was also derived from the explicit observation of atom - water - atom interactions. These findings can be used in improving the protein ？ water interaction models. In the second study, I explicitly considered structural diversity of proteins and ligands, and tested whether that can improve the binding affinity calculation of protein-ligand interaction. This approach was applied to the interaction between SH3 domain and peptides. Consequently, I found that the consideration of the structure ensemble can improve the structure based binding affinity calculation. In addition, several specific issues in ensemble based binding affinity calculation such as the multiple template and sequence to structure alignment problem were approached in the analysis. In the final study, I developed a computational model that evaluates the chemical diversity and accessibility of surface region of the protein. This model was developed to select and design protein molecules that can be used as scaffolds for binding protein library construction. This model was then applied to a set of monomer proteins deposited in the protein structure database and a number of new candidates were selected. The analysis results were compiled into a database that was provides as a web-page. The experimentalist can refer this database when developing a new binding protein scaffold. In addition I developed a web server that performs this analysis to any proteins submitted by the users. During my thesis studies, I analyzed the structure of the protein-water interaction and developed protein models that improved binding affinity calculation and that enable computational selection and design of a binding protein library. These researches would have been expanded our knowledge on the protein structure and I expect that these analyses would contribute to the development of improved computational models of protein molecules.