Disulfide bonding state prediction with SVM based on protein types

Disulfide bonds play the key role for predicting the three-dimensional structure and the function of a protein. In this paper, we propose an algorithm for predicting the disulfide bonding state of each cysteine in a protein sequence. This method is based on the multi-stage framework and the multi-classifier of the support vector machine. We also design a new training strategy to increase the prediction accuracy. It appends the probabilities to the existing features and then starts a new training procedure repeatedly to improve performance. We perform the experiments on the data set derived from the well-known database Protein Data Bank (PDB). We get 94.2% accuracy for predicting disulfide bonding state, which gets improvement 3.5% compared with the previous best result 90.7%.