Determining candidate binding site locations using conserved features

We present a method for representing important conserved features of a protein directly involved in binding. The conserved features are observed by noticing common atoms in proteins that bind a ligand. A method for constructing a fingerprint based upon the geometric configuration of the atoms in the binding region of the protein is presented. The fingerprint can be applied to other protein structural PDB (protein data bank) files to facilitate identification of the specific binding site the fingerprint represents. The algorithm presented complements software developed at the University of Luton, TMSite, and can be used with the ClustalW alignment tool. By matching one amino acid of the fingerprint onto a new protein, then using it as an anchor and searching for the other corresponding amino acids around it gives a simple method to find possible binding sites. The more points of the fingerprint matched the less likely the match is to be coincidental. Thresholds, both geometrical and biochemical, can be altered to adjust sensitivity of the search. This method will be implemented in TMBoundary, a current software project, as an alternative or complementary search method for binding sites in proteins of known structure.