Predicting Protein-Protein Interactions based on ensemble classifiers

Most proteins perform their function by interacting with other proteins. Thus, the research of Protein-Protein Interactions (PPI) is becoming more and more important. A new multi-encoding of sequence method, which represents the sequence of protein more completely, is presented to predict the protein-protein interactions. We train three different support vector systems based on multi-features and then combine three outputs of three different classifiers to vote the optimal result. Result shows that our method can improve predictive ability, moreover, our method show better performance than some of the previously developed methods in present research.