Prediction of Protein-Protein Interaction Based Only on Coding Sequences

Identification of the interactions among proteins is crucial to illustrate their functions, and further, can help to understand the biological processes and provide insights into the mechanisms of diseases. It has became one of the most challenging and important task in the postproteomic researches. Comparing with costly and time-consuming biochemical experiments, the computational methods have attracted much attention due to their low costing and competitive performance. In this paper, we develop a new method based only on coding sequence to identify novel protein-protein interactions (PPIs). To deal with the imbalance problem, we select suitable negative training set before implementing support vector machine. The proposed method is validated on PPIs data of Plasmodium f alciparum and Escherichiacoli, and yields a predictive accuracy of 93.38% and 92.30% respectively. When performed on independent Plasmodium f alciparum and Escherichiacoli datasets, our method displays promising generalization ability.