PREDICTION OF PROTEIN SECONDARY STRUCTURE BY MULTI-MODAL NEURAL NETWORKS

Prediction of protein secondary structure is considered as an important medium step towards determining its three-dimensional structure and function. We have developed Multi-modal Neural Networks (MNNs) to improve the accuracy of the prediction. The MNN employs several sub-networks to predict the secondary structure individually and produce the final result from the outputs of the sub-networks by the majority decision. Moreover, we expand the MNN into a twofold MNN to enhance the prediction ability.