Multi-Label Learning for Prediction of Subcellular Localization of Human Proteins

The prediction of human protein subcellular localization has attracted extensive efforts because it is closely related to development of drugs and basic biology. Especially when the proteins may simultaneously exist in two or more different subcellular locations, the problem becomes more challenging and interesting. The approach proposed in this work integrated the GO (gene ontology) and evolution information of protein to predict the subcellular locations of human proteins with single or multiple sites, covering 14 subcellular locations. Because of novel application patterns of both GO and PSSM, the result is much better than the art of state.