Using Parallel Combined Classifiers to Improve Classification of Proteins

We introduce a novel parallel combined classifiers method for the problems of protein classification and remote homology detection. The method use parallel computing idea to rebuild the high accuracy SVM-based algorithm and the complete coverage nearest neighbor algorithm. We run the two classifiers simultaneously and combine the output result together to reduce the running time and improve the classification accuracy. The remote homology detection experiments based on the SCOP database are presented to show that the parallel combined classifiers outperform all recently presented classifiers. The parallel speedup experiments show that the parallel methods achieved an ideal acceleration effect in share memory mode and message-passing mode.