Improved sequence alignment at low pairwise identity is important for ident
ifying potential remote homologues in database searches and for obtaining a
ccurate alignments as a prelude to modeling structures by homology. Our wor
k is motivated by two observations: structural data provide superior traini
ng examples for developing techniques to improve the alignment of remote ho
mologues; and general substitution patterns for remote homologues differ fr
om those of closely related proteins. We introduce a new set of amino acid
residue interchange matrices built from structural superposition data. Thes
e matrices exploit known structural homology as a means of characterizing t
he effect evolution has on residue-substitution profiles. Given their origi
n, it is not surprising that the individual residue-residue interchange fre
quencies are chemically sensible.
The structural interchange matrices show a significant increase both in pai
rwise alignment accuracy and in functional annotation/fold recognition accu
racy across distantly related sequences. We demonstrate improved pairwise a
lignment by using superpositions of homologous domains extracted from a str
uctural database as a gold standard and go on to show an increase in fold r
ecognition accuracy using a database of homologous fold families. This was
applied to the unassigned open reading frames from the genome of Helicobact
er pylori to identify five matches, two of which are not represented by new
annotations in the sequence databases. In addition, we describe a new cycl
ic permutation strategy to identify distant homologues that experienced gen
e duplication and subsequent deletions. Using this method, we have identifi
ed a potential homologue to one additional previously unassigned open readi
ng frame from the H. pylori genome. (C) 2001 Academic Press.