Research And Improvement In Silico Prediction Model To Identifying Tumor Hla Class ? Neoantigens

Vaccines are important for human health,as they can activate the immune system to prevent or treat infections and other diseases.However,prophylactic cancer vaccines are currently only effective against cancers of viral origin,such as human papillomavirus mediated cervical cancer.Sipuleucel-T is the first therapeutic cancer vaccine approved by Food and Drug Administration(FDA),and it has only a modest effect on the treatment of prostate cancer.Compared with other immunotherapies,such as immune checkpoint blockade(CPB),adoptive cell transfer(ACT),and chimeric antigen induction(CARs),most vaccines cannot cause significant clinical effects.Neoantigen vaccine as an effective cancer vaccine is being developed in recent years,and it is also one of the most cutting-edge anti-tumor vaccine technologies.The inoculation of neoantigen vaccine not only expands the existing neoantigen-specific T cell population,but also induces a wider range of new T cell specificity in cancer patients to enhance tumor inhibition.Meanwhile,the combination of neoantigen vaccine and other immunotherapies,such as PD-1 or PD-L1,is more significant.Therefore,the development of neoantigen vaccine has become the focus of anti-tumor research and development,especially the technology of screening neoantigen peptide has become the key of vaccine preparation.The next-generation sequencing technology has greatly promoted the feasibility of neoantigen vaccine.However,a large number of candidate false positive neoantigen peptides exist in the process from sequencing identifies somatic mutations to T-cell receptor recognition of neoantigen,which brings great trouble to the verification of experiments.This is a difficult hurdle to overcome when designing vaccines against neoantigens.A set of effective and reasonable screening methods is an indispensable part in the preparation of neoantigen vaccine,which has a very high practical significance.This study focuses on the personalized tumor neoantigens prediction research,and summarizes the current progress in tumor neoantigen bioinformatics prediction methods.After collecting real experiment-validated HLA class I epitopes,we considered the physical and chemical properties,molecular structural characteristics,the peptides information entropy combined with HLA index(% rank values)caused by amino acids sequence of peptides.A total of 24 immunogenicity-related characteristics were used for Random forest algorithm to judge whether the epitope can be T-cell recognition and induce immunogenicity.The classification of antigen epitopes has a better performance than previous studies:the optimal AUC(Area Under Curve)of five cross-validation was up to 0.81.In addition,we used external data set to verify our prediction model,and a good performance remained(AUC=0.77).Furthermore,because neoantigen epitopes are from mutated epitopes based on antigen epitopes,and neoantigens produced in the coding region are only one or two amino acids different from the antigenic peptides.Therefore,we added three additional potential immunogenicity-related characteristics to mutation on the basis of antigen epitope prediction model,including:1)Changes in amino acid hydrophobicity before and after mutations;2)Change of%rank before and after mutation;3)Whether the mutation site is located in the anchor position.Finally,the24 relevant antigen prediction features and 3 mutation-related features(27 in total)were used to together train model using random forest algorithm.The optimal five cross-validation AUC was up to 0.78.Based on the two random forest algorithm,we developed a simple web tools,called‘Ineo-Epp'(http://www.biostatistics.online/INeo-Epp/antigen.php),which can be applied to achieve the forecast for neoantigen and antigen peptides immunogenicity further selection,auxiliary shorten neoantigen vaccine preparation process.In this paper,we focused on the process of TCR identification of HLA-peptide complex.Antigens and neoantigen peptides,which was experimentally verified and can generate immune response were collected from literature and relevant databases.We tried to explore the characteristics of producing immunogenic and provided a new method for further screening tumor neoantigen.This work will complement the research on domestic neoantigen prediction in tumors,and help researchers to gain a new understanding of T-cell recognizing antigen peptide,as well as to work in immune research more convenient.