To help elucidate the general rules of globular protein folding, computer simulations of the conformational transition in model proteins having the left-handed, four-helix bundle motif in which the helices are joined by one or two long loops, as in apoferritin and somatotropin, respectively, have been undertaken. In the context of simple tetrahedral lattice protein models, these unique native helix bundle motifs can be obtained by a set of interactions similar to those found in previous simulations of the folding of four-member a-helical bundles with tight bends and fl-barrel proteins including the Greek key motif. The essential features sufficient to produce the four-helix bundle motif with long loops are as follows: (i) a general pattern of hydrophobic and hydrophilic type residues which differentiate the interior from the exterior of the molecule; (ii) the existence of hydrophilic regions in the amino acid sequence that, on the basis of short-range interactions, are indifferent to loop formation but that interact favorably with all the exterior residues of the helix bundle. Thus, these simulations indicate that, to reproduce all varieties of the left-handed four-helix bundle motif, site-specific interactions are not required.The problem of predicting the tertiary structure of a globular protein, given the primary sequence of amino acid residues that constitute it, is one of the most important and most exciting problems in biochemistry, and many attempts recently have been made to solve this problem (1-7). While a wealth of information on the three-dimensional structure of globular proteins exists (5) remarkably Finally, it should be mentioned there are some Monte Carlo simulations of lattice models of globular proteins (19-22), but they suffer from the limitation that the target native state structures are assumed in advance and the possibility of nonnative interactions is ignored.An alternative method of studying protein folding has recently been proposed in the context of simple lattice models having a minimal set of interactions (12,(23)(24)(25)(26)(27). By employing an efficient Monte Carlo algorithm that allows the model chain to hunt over all configurational space, interactions between all spatially close residues are allowed, and the native structure is not assumed a priori. Furthermore, no site-site specific interactions are introduced. That is, all residues of a given type (e.g., hydrophobic) are taken to be the same. From the simulation results on the folding of p-barrels (12,25,26) and a-helical protein models (27), some general rules of folding have begun to emerge. It seems that the minimal sufficient set of conditions required to obtain a unique tertiary structure are as follows: (i) a general hydrophobic/hydrophilic pattern of interactions; (ii) shortrange interactions that marginally favor secondary structure formation (a slight statistical preference for trans conformations in p-sheet proteins and the existence of helical-wheel type interactions in helical proteins, respectively); (i...