Research On Classification Of α Protein Folding Type

Protein folding plays an important role in the field of life science, and folding classification is the basis of protein folding analysis. Protein folding types(PFTs)mainly reflect the core of conservative topology, including the secondary structure units, relative arrangement of secondary structure units and the connection of peptide chains. Folding classification is of great importance in respect of protein topology classification, protein function annotation, ligand binding sites prediction, new compounds screening and drug design. The study includes the following aspects:1. Establishment of template database of α PFTsASTRAL-1.65 database was widely used to study for folding classification, and it is based on SCOP database. 63 folding types whose samples were more than two were selected for the study. According to DSSP database, Template feature parameters(TFPs) reflecting PFT topology were extracted and protein template database was built.2. Establishment of single-template classification methodAccording to TFPs and the results of TM-align structural alignment, scoring function-Fdscore based on feature parameters was built, and the threshold determination method of Fdscore and single-template classification method also were established. In terms of single-template classification method, recognition rate about63 folding types with 804 samples was verified. The results of sensitivity and specificity in experimental set was consistent with that in test set. For the same fold type, the recognition rate of single-template classification method based on Fdscore is higher than that of single-template classification method based on TM-score. For the different types of folding, the recognition rate of the former is slightly weaker than the latter.3. Establishment of multi-template classification methodBased upon protein template database and TFPs, multi-template classification method based on Fdscore function was established. Meanwhile the method was verified and compared with TM-score multi-template classification method. The average sensitivity of Fdscore classification method and TM-score classification method are respectively 71.86% and 69.46%, and the former are 2.4% higher than latter. For the same folding type, the recognition rate of Fdscore classification method is higher than TM-score function. The average specificity of the Fdscore classification method and TM-score classification method are respectively 99.49% and 99.30%, and the results are similar. For the different folding type,the recognition rate of Fdscore classification method and TM-score classification method are similar. The average MCC value of the Fdscore classification method and TM-score classification methodare respectively 0.69 and 0.65, and the former is slightly higher than latter. Overall,Fdscore classification method is better than TM-score classification method.In conclusion, the method of α PTF classification can not only provide the guide for other folding types, but also could classify for unknown protein structures and predict unknown function.