A Protein Structural Classes Prediction Method based on Various Information Fusion

Protein structural class’s knowledge plays an important role in understanding the folding mode of protein. The prediction of protein structural classes as a transitional stage of the secondary structure of the protein to the tertiary structure is considered to be an important and challenging task. In this paper, PSI-BLAST profile is used to extract the evolutionary information of protein, and the position-specific scoring matrix is obtained from PSI-BLAST profile. Then formula is used to transform PSSM into a fixed length feature vector. Extract the protein composition information and sequence order information from the pseudo-amino acid composition, and fuse all the extracted feature vectors. Finally, the fused feature vector is input to the support vector machine classifier to predict protein structural classes. The results were obtained by jackknife test and compared with other prediction methods on the two low similarity benchmark datasets 1189 and 640. The results show that the proposed method can predict the protein structural classes effectively.