To generate structures consistent with both the local and nonlocal interact
ions responsible for protein stability, 3 and 9 residue fragments of known
structures with local sequences similar to the target sequence were assembl
ed into complete tertiary structures using a Monte Carlo simulated annealin
g procedure (Simons et al,, J Mol Biol 1997; 268:209-225), The scoring func
tion used in the simulated annealing procedure consists of sequence-depende
nt terms representing hydrophobic burial and specific pair interactions suc
h as electrostatics and disulfide bonding and sequence-independent terms re
presenting hard sphere packing, alpha-helix and beta-strand packing, and th
e collection of beta-strands in beta-sheets (Simons et al,, Proteins 1999;3
4:82-95), For each of 21 small, ab initio targets, 1,200 final structures m
ere constructed, each the result of 100,000 attempted fragment substitution
s. The five structures submitted for the GASP III experiment mere chosen fr
om the approximately 25 structures with the lowest scores in the broadest m
inima (assessed through the number of structural neighbors; Shortie et al,,
Proc Natl Acad Sci USA 1998;95:1158-1162). The results were encouraging: h
ighlights of the predictions include a 99-residue segment for MarA with an
rmsd of 6.4 Angstrom to the native structure, a 95-residue (full length) pr
ediction for the EH2 domain of EPS15 with an rmsd of 6.0 Angstrom, a 75-res
idue segment of DNAB helicase with an rmsd of 4.7 Angstrom, and a 67-residu
e segment of ribosomal protein L30 with an rmsd of 3.8 Angstrom. These resu
lts suggest that ab initio methods may soon become useful for low-resolutio
n structure prediction for proteins that lack a close homologue of known st
ructure. (C) 1999 Wiley-Liss, Inc.