Mhc Class Ⅱ Affinity Peptide Research Across Subtype Prediction Algorithm

MHC (Major Histocompatibility Complex) molecules play a key role in antigen presentation. Binding of MHC molecules and antigen peptides is a requisite step in antigen presentation of immune responses. Accurate prediction of such peptides that can bind to MHC molecules is of great significance for immunologists and biomedical staff. On the one hand, in theory it can help understand the underlying mechanism of immune recognition and response; on the other hand, in practice it can facilitate T-cell antigen epitope discovery, and then guide the design and development of epitope vaccine to cure many critical diseases.MHC binding peptides is a tiny part of the huge amount of short peptides that originate from antigen protein. Development of computational prediction methods and making predictions in silico are therefore particularly important. MHC has two classes concerning immunity, i.e., class I and II. MHC class II has over one thousand allotypes with different binding specificities. The prediction of MHC class II binding peptides is more challenging for their more variable length, comparing MHC class I. Determination of the binding peptide set of each allotype by biomedical experiments needs lots of money and time and is thus infeasible. On the basis of existing experimentally measured data, development of computational prediction method for MHC class II binding peptides therefore become a research hotspot in Immunoinformatics.A pan-specific method can make predictions for all MHC molecules in principle. To investigate the possibility of further improvement in performance and usability of pan-specific methods, this article extensively reviews existing pan-specific methods and their web servers. We first present a general framework of pan-specific methods. Then, the strategies and performance as well as utilities of web servers are compared. Finally, we discuss strategies and future directions to improve the performance of pan-specific prediction.Base on the previous review and discussion of future direction, we extend TEPITOPE that is extensively used to a pan-specific method based on pocket similarity. The resulting novel method, TEPITOPEpan, can predict binding peptides of over700HLA-DR molecules. TEPITOPEpan has inherited excellent explanation and decent prediction performance from TEPITOPE. It shows better performance in computing efficiency, consistence of binding motif and prediction of nonamer epitope and binding cores, even comparing to the-state-of-art method NetiMHCIIpan.To further improve the prediction performance of MHC class II peptides. ensemble learning approach that integrates existing pan-specific methods by AvgTanh is investigated in this paper. The consensus method, MetaMHCIIpan, whose prediction accuracy is significantly improved, is obtained in the end.What’s more, processing of data from source to validation dataset is specified in this paper and several datasets are then offered. For each proposed method, an online web server is provided for public access. In addition, key points and short summary are given while describing each section, and in the end major deficiencies and future directions are discussed.