| The accuracy of the theoretical description of protein folding and protein interactions is directly related to the accuracy of free energy functions developed for describing biological macromolecules. In particular, free energy of the native conformation should be lower than free energies of arbitrary misfolded models (decoys). Furthermore, in the process of protein folding the computed free energy of the transition state, and its location on the landscape of all protein conformations should correctly reproduce experimentally observed kinetic rates and folding pathways. Finally, biomolecular free energies should reflect the fact that native amino acid sequences are on average energetically optimized for their structures. In this work, free energy functions are developed that are suitable for problems of protein structure prediction, protein-protein docking (interactions across a common interface), folding kinetics and sequence design (both for monomeric proteins and protein-protein complexes). In Chapter 2, an empirical hydrogen bonding potential is presented, and native amino acid recovery profiles are used as a test of its performance. In Chapter 3, this hydrogen bonding potential is employed in the context of a simple physical model of protein folding kinetics. The model is applied to predicting kinetic folding rates and transition state structures on a set of monomeric proteins, and its performance is assessed versus available experimental data. In Chapter 4, the role of electrostatics and hydrogen bonding interactions is investigated with respect to the protein structure prediction and protein docking problems. Extensive decoy sets developed for the purpose of free energy function evaluation are used to gain insight into the usefulness and limitations of continuum electrostatics models in molecular biology. Protein structure prediction and docking problems also provide further tests of the hydrogen bonding potential developed in Chapter 2. Finally, Chapter 5 provides a brief summary of some of our findings, with the emphasis on the limitations of current approaches to biomolecular modeling, and suggests directions of future research based on using ab initio electronic structure methods for studying energetics of biological macromolecules. |
