Combining Physico-chemical Properties with PSSM for Protein Secondary Structure Prediction Using BP Neural Network

A two-stage neural network has been used to predict protein secondary structure based on the method of combining physico-chemical properties of amino acid residues with evolutionary information. We employed CB513 as the dataset. After excluding the protein chains containing X, B and which with sequence length shorter than 30 amino acids, there were 492 protein chains in this dataset totally. The network has been trained and tested by 7-fold cross-validation. The result indicated that the prediction accuracy reached 75.96%, which was 0.5% higher than that of only using PSSM as input. Although Q_H was found to be lower than that of PSSM, C_H had an improvement, which indicates that the method we developed is successful.