Detecting patterns in protein sequences

Abstract

The detection of conserved sequence patterns (motifs) in related proteins often yields valuable structural and functional insights. We describe a method that utilizes rigorous statistics and a depth-first search procedure to efficiently and exhaustively search a set of proteins for significant patterns up to a specified length. Additional procedures classify related patterns into groups and identify protein segments most likely to share a common motif. The utility of the method was demonstrated on several difficult test problems; detection of motifs among 56 proteins in the acyltransferase family, detection of a dinucleotide-binding fold present within a small subset of a set of 91 distantly related and unrelated proteins, detection of the helix-turn-helix motif in 15 distantly related proteins and detection of subtle internal repeats in a prenyltransferase. In a search of a large set of sequences for internal repeats, the method detected novel ankyrin-like repeats in an Escherichia coli protein.