We have designed a distance geometry-based method for obtaining the te
rtiary fold of a protein from a limited number of structure-specific d
istance restraints and the secondary structure assignment. Interresidu
e distances were predicted from patterns of conserved hydrophobic amin
o acids deduced from multiple alignments; A simple model chain represe
nting the protein was then folded by projecting its distance matrix in
to Euclidean spaces with gradually decreasing dimensionality until a f
inal three-dimensional embedding was achieved. Tangled conformations p
roduced by the projection steps were eliminated using a novel filterin
g algorithm. Information on various aspects of protein structure such
as accessibility and chirality was incorporated into the conformation
refinement, increasing the robustness of the algorithm. The method suc
cessfully identified the correct folds of three small proteins from a
small number of restraints, indicating that it could serve as a useful
computational tool in protein structure determination from NMR data.
(C) 1995 Academic Press Limited