Sequence-structure patterns: Discovery and applications

Protein sequence data is being generated at a tremendous rate; however, functional annotation of these proteins is proceeding at a much slower pace. Biologists rely on computational biology and pattern recognition to predict the functionality of proteins. This is based on the fact that proteins that share a similar function often exhibit conserved sequence patterns. Such sequence patterns, or motifs, are derived from multiple sequence alignments and have been collected in databases such as PROSITE, PRINTS, SPAT, and eMOTIF. These patterns help to classify proteins into families where the exact function may or may not be known. Research has shown that these domain signatures often exhibit specific three-dimensional structures. In this paper, we show how starting from a seed sequence pattern from any of the existing sequence pattern databases, and using information from the protein structure databases, it is possible to design biologically meaningful sequencestructure patterns (SSPs). An important by-product of our method to generate sequence-structure patterns is an improved sequence alignment as well as an improved structural alignment of proteins belonging to a family and containing that pattern. Validation was performed by matching the resulting SSPs to domains in the ASTRAL compendium associated with a family or super-family designation in the SCOP database. SSPs generated by this method were frequently either fully specific (no false positives), fully sensitive (no false negatives), or both (diagnostic).