Files
Abstract
Carbohydrates play a pivotal role in various life processes including energy metabolism, storage, immune recognition, transportation, signaling and biosynthesis. In these roles, they often interact with other integral components of the living system such as proteins and lipids. An understanding of how these molecules interact can further our knowledge of crucial biological processes, and begins with the knowledge of the three-dimensional structures of these complexes. However, owing to challenges involved in crystallizing oligosaccharide structures, theoretical modeling methods such as molecular docking are often used to predict how oligosaccharides interact with protein receptors. But, docking programs have generalized scoring functions which often produce unnatural oligosaccharide conformations during docking. In this thesis, we present two approaches to improve protein-carbohydrate docking by accounting for specific intra- and intermolecular interaction energies relating to carbohydrates, which are not currently dealt with by existing docking methodologies. In the first approach, we developed a set of Carbohydrate Intrinsic (CHI) energy functions in order to account for intramolecular energies of carbohydrate ligands primarily determined by the conformations of glycosidic torsion angles connecting individual saccharides. This work resulted in the development of Vina-Carb (incorporation of the CHI energy functions within the scoring function of AutoDock Vina), which significantly improved the conformations of oligosaccharide binding mode predictions. In the second approach, we developed a scoring function by fitting a mathematical model to data from literature describing the energy contributed by CH/ interactions. This energy function was used to score the crucial interactions between CH groups lining the carbohydrate ring and the electron densities in aromatic amino acids of interacting proteins. Employing the CH/ interaction energy function to rescore docked protein-carbohydrate complexes improved the rankings of accurate pose predictions made by both AutoDock Vina and Vina-Carb. The scoring functions developed and used in this work are transferable and can therefore be used with other docking programs and also in the refinement of experimental carbohydrate structures.