(38) Simon Sherman, Stanley L. Sclove, Oleg Shats and Leonid Kirnarsky. Dihedral Probability Cluster Monte Carlo Procedure for Conformational Analysis of Proteins. Internet Journal of Chemistry, Vol. 1 (1998), Article 106, 8 pp. URL: http://www.ijc.com/articles/1998v1/106.

A mixture-model cluster analysis of dihedral angle distributions observed in high-resolution proteins from the Protein Data Bank has shown that conformation of the amino acid residues are very highly clustered in the space of the (phi, psi, chi1) dihedral angles. Based on these results, an effective new procedure (called DPC-MC, dihedral probability cluster Monte Carlo) for sampling the conformational space of proteins was developed. DPC-MC uses the cluster probabilities (relative frequencies) to prioritize the sampling of conformational space. Means and standard deviations for the torsion angles are used in this procedure to estimate values of the angles and the angular ranges. DPC-MC may be incorporated into any protocol for protein structure determination and analysis using values and/or allowed ranges for dihedral angles as input data. The procedure allows one to generate "good" intial conformations of proteins in contrast to randomly chosen ones. DPC-MC can be beneficial for three-dimensional protein structure determination from NMR data, building protein conformations from C-alpha-atom traces, homology modeling of proteins, de novo predicting of polypeptide conformations, etc. The benefits of using the DPC-MC procedure were demonstrated by reconstructing complete three-dimensional structures of a protein, flavodoxin, from its C-alpha-atomic coordinates.

KEY WORDS: Monte Carlo, Markov chain, protein structure