HIERARCHICAL ALGORITHM FOR COMPUTER MODELING OF PROTEIN TERTIARY STRUCTURE - FOLDING OF MYOGLOBIN TO 6.2 ANGSTROM RESOLUTION

We present a method for determining the tertiary structure of alpha-pr oteins through computer simulation. A simple model is used in which re sidues are described by a discrete set of phi-psi dihedral angle confo rmations and the extent of alpha-helical regions is assumed to be know n and held fixed. The algorithm makes use of the assigned secondary st ructure to construct a representation of the molecule where helices an d loops are described by cylinders and spheres respectively. Correspon dence with the residue conformations is maintained by using loop geome tries from a list of structures calculated using the allowed dihedral angles in segments of the appropriate length. The optimization consist s of a Monte Carlo simulated annealing procedure combined with a genet ic algorithm in which additional structures are generated by combining parts of different members of the ensemble. We present as an example the results of simulations carried out on myoglobin in which we obtain a low-energy structure with root mean square deviation of 6.2 Angstro m from the native structure. The inter-residue distance map shows that all regions of helix-helix contact are represented confirming that th e overall topology of the folded structure is correct. The root mean s quare deviation of the helical regions alone is 4.1 Angstrom, indicati ng that the packing of the helices is well-reproduced by the present m odel.