Source:Journal of Chemical Physics, 97:9412–9426, 1992
A discretized model of globular proteins is employed in a Monte Carlo study of the helix–coil transition of polyalanine and the collapse transition of polyvaline. The present lattice realization permits real protein crystal structures to be represented at the level of 1 Å resolution. Furthermore, the Monte Carlo dynamic scheme is capable of moving elements of assembled secondary and supersecondary structure. The potentials of mean force for the interactions are constructed from the statistics of a set of high resolution x‐ray structures of nonhomologous proteins. The cooperativity of formation of ordered structures is found to be larger when the major contributions to the conformational energy of the low temperature states come from hydrogen bonds and short range conformational propensities. The secondary structure seen in the folded state is the result of an interplay between the short and long range interactions. Compactness itself, driven by long range, nonspecific interactions, seems to be insufficient to generate any appreciable secondary structure. A detailed examination of the dynamics of highly helical model proteins demonstrates that all elements of secondary structure are mobile in the present algorithm, and thus the folding pathways do not depend on the use of a lattice approximation. Possible applications of the present model to the prediction of protein 3D structures are briefly discussed.