Computational insights into protein-ligand interactions

This thesis describes the use of computational methods as a method of gaining greater insight into protein-ligand interactions, both at the molecular-recognition level and as an aid to mechanistic interpretation. Chapter Two:Part One addresses one of the major approximations in molecular docking, the rigid protein approximation. A “cross-docking” study was conducted using 31 neuraminidase complexes. It was shown that some protein conformations are more able to bind a wide variety of ligands than others but that it is possible to obtain a correct binding mode even if small conformational changes to accommodate the ligand are observed in the actual crystal complex. Chapter Two:Part Two considers a water-scoring algorithm that allows selected water molecules to be included if they are beneficial to a docked conformation. This methodology was first applied to three key mediating waters observed in neuraminidase complexes and then to a large validation set with false waters added. The prediction of water occupancy was very good (~90%) and only a small reduction in the quality of predictions is observed. A number of mechanisms for ketopantoate hydroxymethyltransferase (KPHMT) are proposed in Chapter Three:Part One. Through use of molecular docking, the structural and chemical data are rationalised to propose a mechanism consistent with observed results. In Chapter Three:Part Two a virtual screen of a large targeted library of compounds was conducted and a number of potential KPHMT binding molecules proposed. Chapter Four involves the creation of a closed model for E. coli pantothenate synthetase. The model was then used for molecular docking of the reaction intermediates, to give insight into the mechanism, and potential inhibitors.EThOS - Electronic Theses Online ServiceGBUnited Kingdo