Prediction of protein interactions by combining genetic algorithm with SVM method

This paper proposes a novel hybrid GA/SVM method that can predict the interactions between proteins intermediated by the protein-domain relations. Firstly, we represented a protein by the domains contained inside, which can consider the effects of domain duplication. To simulate the combination of different domains, a transformation of the domain composition was taken subsequently. Further, a genetic algorithm was used to seek the optimized transformation, which had been adopted as the input vector of a predictor constructed using support vector machines method. Finally, experiment results validated the effectiveness and efficiency of our proposed approach by a better performance.