A SVM-based system for predicting protein-protein interactions using a novel representation of protein sequences

Protein-protein interactions (PPIs) are crucial for almost all cellular processes, including metabolic cycles, DNA transcription and replication, and signaling cascades. However, the experimental methods for identifying PPIs are both time-consuming and expensive. Therefore, it is important to develop computational approaches for predicting PPIs. In this article, a sequence-based method is developed by combining a novel feature representation using binary coding and Support Vector Machine (SVM). The binary-coding-based descriptors account for the interactions between residues a certain distance apart in the protein sequence, thus this method adequately takes the neighboring effect into account and mine interaction information from the continuous and discontinuous amino acids segments at the same time. When performed on the PPI data of Saccharomyces cerevisiae, the proposed method achieved 86.93% prediction accuracy with 86.99% sensitivity at the precision of 86.90%. Extensive experiments are performed to compare our method with the existing sequence-based method. Achieved results show that the proposed approach is very promising for predicting PPI, so it can be a useful supplementary tool for future proteomics studies.