Institute for Chemical Research, Kyoto Univ.
Recently we have construct the Enzyme-Reaction Database which links amino acid sequences of enzymes to chemical structures recognized as their ligand. For each enzyme, sequence were clustered according to their sequence similarity, and aligned in eachcluster to extract a conversed sequence. Conversed sequence was cut into fragments of five-residue long, which are sequence fragments relating an enzyme to its ligande and stored on a relational database management system, SYBASE, with other items in the Enzyme-Reaction Database. To infer the biological function of a query sequence, we constructed a system that searches the fragment with the query sequence by using SQL. This system outputs a list of ligands and enzyme names related by the fragments found in the query sequence. When a sequence of glutathione synthetase was a query against 99,535 fragments, 1,697 fragments were hit allowing one mismatch in five residues. Most fragments were derived from ATP-related enzymes and related to HO, ATP, and O as ligands of the query enzyme. Because ligands output by the present system are biased to ATP and its related compounds, it is difficult to identify the biological function of the query sequence. This bias reflects that of the enzymes and ligands registered in the sequence database.