A Neural Network to Locate the Copper-binding sites of Metalloprotein

This paper presents an application of neural networks in location of the copper-binding sites of metalloprotein. Using annotated metalloprotein downloaded from PDB, sequences including copper-binding sites were extracted. By further finding the particular core segments of copper-binding sites, the input and output information for training is polished. Moreover, this paper investigates the number of input nodes whose input information was coded by the residues´ hydrophobic values, the number of hidden nodes, the size of training window. Back propagation algorithms were chosen for training neural networks. Results showed that the method was capable of efficiently identifying copper-binding proteins and predicting copper-binding sites at a very high accuracy