Protein backbone dihedral angle prediction based on probabilistic models

Protein backbone dihedral angles are important descriptors of local conformation for amino acids. Protein backbone dihedral angle prediction lays the foundation for prediction of higher-order protein structure. Existing prediction methods of protein backbone angles mainly exploit traditional machine learning techniques. In this paper, we propose to use two well-known types of probabilistic models - maximum entropy Markov models (MEMMs) and conditional random fields (CRFs) to predict the backbone dihedral angles of amino acid sequences. Experiments conducted on dataset PDB25 show that these two probabilistic models are effective in dihedral angle prediction, and CRFs outperform MEMMs.