Utilizing NMR spectroscopy and molecular docking as tools for the structural determination and functional annotation of proteins

With the completion of the Human Genome Project in 2001 and the subsequent explosion of organisms with sequenced genomes, we are now aware of nearly 28 million proteins. Determining the role of each of these proteins is essential to our understanding of biology and the development of medical advances. Unfortunately, the experimental approaches to determine protein function are too slow to investigate every protein. Bioinformatics approaches, such as sequence and structure homology, have helped to annotate the functions of many similar proteins. However, despite these computational approaches, approximately 40% of proteins still have no known function. Alleviating this deficit will require high-throughput methods that combine experimental and computational approaches. Nuclear magnetic resonance (NMR) ligand affinity screens are an experimental approach that can detect protein-ligand interactions, measure a corresponding dissociation constant, and reliably identify the ligand binding site. Correspondingly, molecular docking is a computational tool that can be used predict the location of the binding site and conformation of a compound when bound to a protein using only the structures of both the protein and the compound. Molecular docking provides a rapid way to generate protein-ligand costructures and evaluate numerous compounds in a large chemical library. Together, molecular docking and NMR ligand affinity screens provide valuable information for determining the function of a protein. This dissertation describes the high-throughput application of the Functional Annotation Screening Technology by NMR (FAST-NMR), which combines NMR ligand affinity screens, molecular docking, and bioinformatics to help determine the function of 20 previously uncharacterized proteins. Additionally, new tools were developed to utilize 2D 1H, 15N-HSQC (heteronuclear single quantum coherence) chemical shift perturbations (CSPs) and molecular docking to generate consensus binding sites (CSP-Consensus) and protein-ligand costructures (AutoDockFilter). Virtual screening was also successful utilized to identify a potential natural ligand and propose a function for the YndB protein from Bacillus subtilis. Finally, the solution structure of human protein DNAJA1 was determined and its potential role in pancreatic cancer investigated