Prediction of Protein-Protein Interactions Using Local Description of Amino Acid Sequence

Protein-protein interactions (PPIs) are essential to most biological processes. Although high-throughput technologies have generated a large amount of PPI data for a variety of organisms, the interactome is still far from complete. So many computational methods based on machine learning have already been widely used in the prediction of PPIs. However, a major drawback of most existing methods is that they need the prior information of the protein pairs such as protein homology information. In this paper, we present an approach for PPI prediction using only the information of protein sequence. This approach is developed by combing a novel representation of local protein sequence descriptors and support vector machine (SVM). Local descriptors account for the interactions between sequentially distant but spatially close amino acid residues, so this method can adequately capture multiple overlapping continuous and discontinuous binding patterns within a protein sequence.