Folding of globular proteins by energy minimization and Monte Carlo simulations with hydrophobic surface area potentials

We describe an efficient method for the analytical calculation of the solvent accessible surface areas and their gradients in proteins on serial and parallel computers. We applied energy minimizations and Monte Carlo simulations to the small three-helix bundle protein Er-10. The force field consisted of the ECEPP/2 energies and a term describing protein-solvent interaction through the solvent accessible surface area. We show that the NMR structure is stable when refined with this force field, but large structural changes are observed in energy minimization in vacuo. When we started from random tertiary structures with preformed helices and maintained the helical segments by dihedral angle constraints, the final structures with the lowest energies resembled the native fold. The root-mean-square deviations for the backbone atoms of the three helices compared to the experimentally determined structure were 3 Angstrom to 4 Angstrom.