The effective feature representations by integrating PseAAC and PSSM for protein sub-nuclear location

In molecular biology, many effective representations have been used for sub-nuclear location. However, these representations, containing only single characterize of protein, make the information for classification insufficient. Inspired by this fact, we propose two integrated representations by feature fusion method in the expectation of extracting richer information from the protein sequence. Concretely, we combine pseudo amino acid composition (PseAAC) with position specific scoring matrix (PSSM) and then obtain two fused representations, called briefly PseAACGreyPSSM and PseAACPsePSSM. Both the proposed representations, thus, include the protein amphiphilic factors and the biological evolution information. We, with feature vectors of protein samples, adopt the K nearest neighbor (KNN) classifier for predicting. The final experimental results state that our proposed representations are superior to the single feature representations consistently.