Author Description

Protein - protein interaction play a major part in Bioinformatics and Computational Biology to interpret the basic inherent principles of biological organizations. Compared to the available protein networks and sequences of dissimilar species, the number of exposed PPIs (protein-protein interactions) is still very much confined. In this book, a sequence-based method using the physio-chemical properties of protein residues is proposed by blending a new feature representation Auto-Covariance (AC) and Artificial Neural Network (ANN). Auto-Covariance describes for interactions among amino acids residues at some outstrip apart in the primary protein sequence, so this method takes the adjacent effect into account for protein interactions. The amino acids were changed into numerical values presenting six physico-chemical properties, after that those numerical sequences were changed to one dimensional vector of same size by AC. Finally, the ANN model was build using the vectors of Auto-Covariance variables as input. The experiment demonstrated that the prediction model and the interaction prediction accuracy is higher than 94.0% after 5 fold cross validation.