The design of a ligand discovery and optimization system for structure-based drug design

Unprecedented opportunities in drug design now exist because the basis of disease is being understood on a more molecular level. With the increasing number of known protein structures, it is likely that the structure of the target enzyme will be known or can be determined using X-ray crystallography, NMR techniques, or homology modeling. The structure of the enzyme can be used to design new therapeutic agents that are complementary to the receptor and bind with high affinity. In this dissertation I have developed a new system for the discovery and optimization of new lead compounds in the drug design process. The system contains a molecular graphics component as well as integrated molecular docking methods. Molecular graphics are important in the drug design process. They identify receptor-ligand interactions and suggest modifications to the lead compound that may improve binding. The system uses the most recent developments in computer graphics to interactively compute and display molecular surfaces and volumes of interest. Texture mapping is also used to allow the chemist to more clearly render the information. Molecular docking methods are invaluable for drug design. They are used to screen for new lead compounds and identify chemically plausible structures for the intermolecular complex. Once a binding mode is found, chemical modifications to the lead compound can be suggested. In this dissertation I developed methods which use genetic algorithms to dock flexible ligands to rigid receptors. I studied the use of molecular mechanics force fields and empirical estimates of the binding affinity as scoring functions. I found that the latter work much better within the context of a genetic algorithm optimization. I performed three different types simulations on seven different receptor-ligand complexes. With an initially docked fragment, my system performed as well as other incremental construction approaches. I also demonstrate that my system is robust to errors in the orientation of the base fragment. Furthermore, in all cases, when no information about the binding was provided, my system found solutions very close to the crystal structure. This system should prove to be useful for the design of new therapeutic agents.