Hierarchical modeling method of protein side chain prediction

The invention provides a hierarchical modeling method of protein side chain prediction. The hierarchical modeling method of the protein side chain prediction comprises the following steps: (1) taking main chain information as input, executing a first layer reasoning unit, outputting a side chain torsion angle X1, (2) taking the main chain information and the side chain torsion angle X1 as input, executing a second layer reasoning unit, outputting a side chain torsion angle X2, (3) taking the main chain information, the side chain torsion angle X1 and the side chain torsion angle X2 as input, executing a third layer reasoning unit, outputting a side chain torsion angle X3, and (4) taking the main chain information, the side chain torsion angle X1, the side chain torsion angle X2 and the side chain torsion angle X3 as input, and executing a forth layer reasoning unit, outputting a side chain torsion angle X4. The invention further provides a practice process of the hierarchical modeling method for all the layer reasoning units based on the hierarchical modeling method. Meanwhile, for all the layer reasoning units and combining data bank network (DBN) models corresponding to all the layer reasoning units output by the practice process, the invention further provides a sampling process of the hierarchical modeling method.

[1]  Roland L. Dunbrack,et al.  Bayesian statistical analysis of protein side‐chain rotamer preferences , 1997, Protein science : a publication of the Protein Society.