Research On Contact Map Based Protein Structure Prediction Method

The scale of the protein sequence database has been growing rapidly.However,due to the limitation of experimental techniques,the tertiary structures of many protein sequences remain unknown.To solve this problem,many structure prediction methods have been developed in the past decades.Recently,the increasing accuracy in protein contact map prediction opens a new avenue to improve the performance of structure prediction algorithms.We have developed a new template-based structure prediction algorithm named CATHER,which combines both conventional sequence profile and contact map predicted by a deep learning-based algorithm,and ranks the results of template selection by all alignment results on the template library.Benchmark on an independent test set and CASP12 targets demonstrated that CATHER made significant improvement over other method on hard targets.Besides,the research on the contribution of the contact map indicates that the contact information refines not only the alignment of the query and the template,but also the ranking score of the alignment,and the more accurate the predicted contact map is,the greater the improvement will be.Our method ranked at the top 10 among all 39 participated server groups on the 32 free modeling targets in the blind tests of the CASP13 experiments.Based on the structure prediction method mentioned above,we built a protein structure prediction server CATHER.Users of the server can get the prediction results by submitting the query sequence,and view the predicted structure,contact map,secondary structure,relative solvent accessibility and so on.The server is available at: http://yanglab.nankai.edn.cn/CATHER.