Loop problem in proteins: Developments on Monte Carlo simulated annealing approach

Calculations of loop segments in bovine pancreatic trypsin inhibitor starting from random conformations are more efficient, reproducible, and reliable due to several program enhancements. Monte Carlo simulated annealing (MCSA) calculations of a five-residue a-helix N-terminus segment (H5) and @-strand segment (B5) in this study are compared to the corresponding loop calculations in our previous study. Characteristics of the calculations are: the lowest final total energy conformations (LECs) are within 5 kcal/mol; the average backbone deviations of the computed segments from the native X-ray conformations are 0.43 -t 0.15 A for H5 and 0.68 -t 0.20 A for B5; and all the native backbone-backbone hydrogen bonds (H bonds) are present in the best LECs. Compared to the previous study, the H5 and B5 calculations are about 3 and 24 times more efficient, respectively. In the analysis of the best H5 simulated annealing run, backbone-backbone H bonds appear between RT = 4 and 70 kcal/mol, where RT is the Boltzmann temperature factor. H bonds that involve side chains appear in the RT = 1-10 kcal/mol range. Program enhancements implemented are varying main chain versus side chain dihedral angle selection rate, varying +/I,!J and xl/x2 dihedral angles in pairs, the use of main chain and side chain rotamer libraries, and varying the location of the segment origin.