Pairwise iterative superposition of distantly related proteins and assessment of the significance of 3-D structural similarity.

A challenge lies in identifying distant protein 3-D structural similarity by rigid-body superposition. The most common measure of structural similarity is r.m.s. distance (r.m.s.d.) between topologically equivalent residues, and most automated methods of protein modelling rely on the assembly of rigid fragments from known 3-D structures. A fast method of improving the definition of a common protein fold by superposition, especially for distant relationships, is described. The definition of topological equivalence by the standard dynamic programming sequence alignment algorithm is extended by refining the entire structure alignment (not just those equivalenced residues within a given cut-off distance) and determining whether the alignment can be continued at the termini. The most appropriate distance-based definition of topological equivalence for a given comparison is identified. Despite the fact that hitherto the distant similarity between the globin fold and colicin A has not been recognized directly by rigid-body superposition, this new approach defines more equivalent residues with a lower r.m.s.d. between them than that obtained by the superposition of equivalences identified by a more elaborate method. A previous distance metric of 3-D structural similarity derived from rigid-body superposition has been extended to the assessment of superpositions where topological equivalences have been determined by methods other than rigid-body ones.