A novel approach has been taken to the problem of protein folding that examines the complete range of folded topologies accessible in the compact state of globular proteins. The procedure is to generate all conformations, with volume exclusion, upon a lattice in a space restricted to the individual protein's known compact conformational space. This approach affords an enormous reduction in the number of conformations that must be considered for the complete problem. Five proteins have been treated using this method; avian pancreatic polypeptide (36 residues), crambin (46 residues), rubredoxin (52 residues), pancreatic trypsin inhibitor (58 residues), and neurotoxin (62 residues). All conformations generated are evaluated in terms of residue-specific, pairwise contact energies that favor non-bonded, hydrophobic interactions. Native structures for these proteins are always found within the best 3% of all conformers generated. Additional methods are being developed to further narrow the choice of the best among the good conformers. One approach superimposes all atoms onto a lattice conformation, and refines this structure with conventional molecular dynamics and energy minimization. For the test case of one low energy lattice conformation for crambin, an all-atom structure was built that was within 1.46 A RMS deviation from the native crystal structure. These methods are simple and general and can be used to determine most favorable overall packing arrangements for the folding of specific amino acid sequences within a restricted space. Other aspects being considered include: folding intermediates, combining the folding calculations with sequence homologies and investigating hydrophobic cores, binding, choices of overall shapes and the relationship between good packing and secondary structures. The residue potentials are being generalized for application to situations with differing amounts of solvent interactions.
