Abstract
This paper proposes a neural network approach to estimate the enzyme active sites, irrespective of homology, especially the catalytic triads located in chymotrypsin sequences. The three-layered network model is trained by all functional classes of enzymes. The model feeds amino acid residue sequences and outputs the classification for each residue. Outputs are compared with the pre-determined catalytic triad locations of the typical chymotrypsin. It was observed that the triads tend to locate in a valley between two successive peaks of the output. This estimation was tested using location-known chymotrypsins and detected 21/22 correct residues. Further studies are needed, however, this will have utility in finding unknown active sites.