A Four-Layer Reference Model for Prediction of Protein Side-Chains Based on Dynamic Bayesian Networks

A four-layer reference model for prediction of protein side-chains based on Dynamic Bayesian Networks is proposed. The relation of backbone with four side chain torsion angles is considered in this model, and the problem caused by sparse data is resolved to a certain extent. Experiment shows that this method has produced a better DBN model and more exact prediction results. It indicates that the four-layer reference model is effective.