Prediction of Protein Subcellular Multi-localization by Using a Min-Max Modular Support Vector Machine

Prediction of protein subcellular location is an important issue in computational biology because it provides important clues for characterization of protein function. Currently, much effort has been dedicated to developing automatic prediction tools. However, most of them focus on mono-locational proteins. It should be noted that many proteins bear multi-locational characteristics, and they carry out crucial functions in biological processes. This work aims to develop a general pattern classifier for predicting multiple subcellular locations of proteins. We used an ensemble classifier, called min-max modular support vector machine (M-SVM), to solve protein subcellular multi-localization problem, and proposed a task decomposition method based on gene ontology (GO) semantic information for the M-SVM. We applied our method to a high-quality multi-locational protein data set. The M-SVMs showed better performance than traditional SVMs using the same feature vectors. And the GO decomposition also helped improve the prediction accuracy with more stable performance than random decomposition.