Exploring The Molecular Design Of Ligand Binding Sites By Computational Protein Design

Ligand binding sites in natural proteins, with diverse structural details, provide the foundation for enzymatic activity, antibody-antigen recognition, ligand-induced pathway activation and drug discovery in general. The work presented in this dissertation seeks to understand the general design principles of the molecular details revealed in the ligand-protein complex structures. An engineering approach based on computational protein design was taken to expand the boundary of our current knowledge. By combining computational structural modeling and protein biochemical characterization, computational design of ligand binding proteins iterates between structure-based design hypotheses and experimental validation. This research scheme was applied to two related topics: 1) re-purposing natural ligand binding sites and 2) designing de novo ligand binding proteins. Representative small molecules, steroids (digoxigenin, 17-hydroxylprogesterone, cortisol) and an environmentally sensitive fluorophore (DFHBI), were chosen as design targets. High-resolution X-ray crystal structures of the engineered proteins were obtained and analyzed for modeling feedback. Binding affinity and specificity, protein stability and function, as well as modeling challenges were discussed in each case. The design methods developed and tested in this work represent a systematic way of engineering small molecule binding sites and can be expanded to broad applications. As a rigorous test of our current knowledge, computational design of ligand-binding proteins presented in this work emphasizes the high precision required for accurate ligand positioning and protein conformation modeling.