Protein folding at atomic resolution: Potentials, dynamics, and packing

An all-atom Monte Carlo simulation of protein folding is developed and tested on the α/β protein crambin. Ensemble folding kinetics are described. It is shown that with a sufficiently stabilizing potential, folding to less than 1 Å deviation from crystal structure coordinates is observed. Inclusion of correct side-chain geometry leads to complex behavior during simulation, including slow dynamics after native topology is obtained, and modified thermodynamics. The model is applied to study the side-chain packing problem for a fixed native backbone in large proteins. The conformational space of side-chains is fully characterized using umbrella sampling. We find that 95% of positions can accommodate more than one rotamer and that the entire protein can be repacked in a huge number of ways, without steric clashes.; Protein folding simulations are then re-examined to address side-chain ordering. A simple lattice model of folding is introduced in which side-chain rotamers are modeled as spins. Constrained dynamics of spins is introduced to mimic excluded volume. The model is found to exhibit glass-like dynamics due to side-chain degrees of freedom. Side-chain relaxation times are calculated, and fastest freezing residues are shown to be located at nucleus positions. Thermodynamics are analytically derived, and relevant experiments are discussed.; A method is presented for derivation of all-atom, sequence-based protein folding potentials. The method is tested on a β-hairpin, an α-helix and a small protein. Protein folding to 2.5 Å deviation from native coordinates is achieved with these potentials. Potentials derived for various proteins are compared.; The number of sequences folding into a given protein structure is called the designability of the structure, and is thought to be a relevant factor governing the kinds of protein structures that are observed in living organisms, and their evolution. Designability is studied analytically in protein-like square lattice structures. Structural features enhancing designability are identified using a statistical mechanics approach. These features are shown to be sensitive to details of the potential of interaction. The approach provides an explanation of the observation that designability is highly variable across proteins, both on-lattice and in real life.