A common approach to protein modeling is to propose a backbone structu
re based on homology or threading and then to attempt to build side ch
ains onto this backbone, A fast algorithm using the simple criteria of
atomic overlap and overall rotamer probability is proposed for this p
urpose, The method was first tested in the context of exhaustive searc
hes of side chain configuration space in protein cores and was then ap
plied to all side chains in 49 proteins of known structure, using simu
lated annealing to sample space, The latter procedure obtains the corr
ect rotamer for 57% and the correct chi(1) value for 74% of the 6751 r
esidues in the sample, When low-temperature Monte-Carlo simulations ar
e initiated from the results of the simulated-annealing processes, con
sensus configurations are obtained which exhibit slightly more accurat
e predictions, The Monte-Carlo procedure also allows converged side ch
ain entropies to be calculated for all residues. These prove to be acc
urate indicators of prediction reliability, For example, the correct r
otamer is obtained for 79% and the correct chi(1) value is obtained fo
r 84% of the half of the sample residues exhibiting the lowest entropi
es, Side chain entropy and predictability are nearly completely uncorr
elated with solvent-accessible area, Some precedents for and implicati
ons of this observation are discussed. (C) 1996 Wiley-Liss, Inc.