Prediction of catalytic site of proteins based on amino acid triads approach using non parametric function

Abstract

In this study, we present a method for the catalytic site prediction of proteins rest on the triads of amino acids residues using non parametric function - artificial neural network. Using this method, we can efficiently predict that whether the amino acid triads of a protein are the part of catalytic site or not. For the preparation of training and test datasets, catalytic site residues of protein are downloaded from the database of catalytic site atlas and residues for non catalytic site are taken which are not participating in the formation of catalytic site of protein. This method used the numerical value of six physiochemical properties of amino acids along with the difference between centers of mass of whole protein and amino acids triads as the input for the neural network. Our analysis shows that this method is worked with the efficiency of 83.66% which is higher than other existing model for the prediction of catalytic site of protein. Our analysis is based on the residues physiochemical and topological properties and not on the evolutionary and sequence similarities so, In future, this work may help the researchers to develop tool and predicting the nature of residues of catalytic or active site of protein and may be helpful in ligand designing and molecular docking.