https://orcid.org/0000-0003-1359-8024
KEYWORDS:
Neoantigens are peptides encoded by mutations in specific tumors and delivered to the surface of tumor cells by complex core adaptation proteins (MHCs), where they can be described as “foreign” by T cells insider to activate an immune response.Citation1 Prediction of neoantigens is effective in the development of cancer vaccines. Studies refer to the development of a personalized vaccine to treat gastrointestinal cancers, uterine cancer, bladder cancer, pancreas, breast cancer, melanoma and glioblastoma.Citation2,Citation3
Due to technical problems and costs, the prediction of antigens presentation by human leukocyte antigen (HLA) with intelligent machine-learning tools can be effective approaches. Artificial neural network is one of the models used in machine learning and deep learning.Citation2 These models are used to predict the immune response to a vaccine, evaluate the effectiveness of vaccines in the cellular and molecular domains, predict HLA binding and identify the best peptide for each patient’s individual vaccines.Citation4 Python,Citation5 PERL,Citation6 R and MATLABCitation7 are used in these computational predictions and deep learning strategies.
The use of integrated models and architectures increases the accuracy of predictions. One of these models is MARIA (major histocompatibility complex analysis with recurrent integrated architecture). MARIA helps to better detect immunogenic epitopes in diverse cancers and autoimmune disease.Citation2 Therefore, identifying and combining machine learning models and creating integrated architectures for accurate and more comprehensive prediction is an effective step. In formulating the framework, these components of multi biological data, multi features extraction, multi feature selection, multi classifier and algorithms, multi cross validations and final architecture should be considered. Combining several machine learning algorithms and their integration models seems necessary. Integrated algorithms improve predicting accuracy and effectiveness.
One of the challenges is complex variables detection. Vaccine safety and efficacy data, recurrence after vaccination, clinical data, signs and symptoms of different types of cancer, tissue data and blood samples are included in this complex. HLA typing, mutated peptides, HLA binding, candidate neoantigen prioritization and selection are examples of biological data.Citation4 A specialized team is needed to confirm the comprehensiveness and completeness of the data. Oncologists, geneticists and information managers can play a key role at this stage. Pre-processing feature selection can facilitate via SNOMED-CT (Systematized Nomenclature of Medicine – Clinical Terms) integrated with ICD11 (International Classification of Diseases 11th Revision) and natural language processing. TEPITOPEpan, PREDIVAC, Propred, EpiMatrix, EpiDock, EpiTOP and EpiPredict are examples of databases where cancer vaccine data can be extracted. In the stage of multi features extraction, selecting the appropriate tools and platform for data extraction is important. These platforms can be commercial or open source. Tools and data types and tasks must be carefully analyzed. The combination of algorithms such as QM (quantitative matrix), ANN (artificial neural network), SVM (support vector machine) and BIMAS (quantitative matrix-based tool)Citation8 can be used in this architecture. Develop or optimize computational approaches to predicting immunogenic variables and the development Integrated architectures in this field are effective in achieving positive clinical outcomes for cancer patients.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Additional information
Funding
References
- Tran NH, Qiao R, Xin L, Chen X, Shan B, Li M. Personalized deep learning of individual immunopeptidomes to identify neoantigens for cancer vaccines. Nature Mach Intell. 2020;2(12):764–71. doi:10.1038/s42256-020-00260-4.
- Chen B, Khodadoust MS, Olsson N, Wagar LE, Fast E, Liu CL, Muftuoglu Y, Sworder BJ, Diehn M, Levy R, et al. Predicting HLA class II antigen presentation through integrated deep learning. Nat Biotechnol. 2019;37(11):1332–43. doi:10.1038/s41587-019-0280-2
- Kiyotani K, Toyoshima Y, Nakamura Y. Immunogenomics in personalized cancer treatments. J Hum Genet. 2021;1–7.
- De Mattos-Arruda L, Vazquez M, Finotello F, Lepore R, Porta E, Hundal J, Amengual-Rigo P, Ng CKY, Valencia A, Carrillo J, et al. Neoantigen prediction and computational perspectives towards clinical benefit: recommendations from the ESMO precision medicine working group. Ann Oncol. 2020;31(8):978–90. doi:10.1016/j.annonc.2020.05.008.
- Phloyphisut P, Pornputtapong N, Sriswasdi S, Chuangsuwanich E. MHCSeqNet: a deep neural network model for universal MHC binding prediction. BMC Bioinform. 2019;20(1):1–10. doi:10.1186/s12859-019-2892-4.
- Kaushik AC, Li M, Mehmood A, Dai X, Wei D-Q ACPS. An accurate bioinformatics tool for precision-based anti-cancer peptide generation via omics data. Chem Biol Drug Design. 2021;97(2):372–82. doi:10.1111/cbdd.13789.
- Yamankurt G, Berns EJ, Xue A, Lee A, Bagheri N, Mrksich M, Mirkin CA. Exploration of the nanomedicine-design space with high-throughput screening and machine learning. Nature Biomed Eng. 2019;3(4):318–27. doi:10.1038/s41551-019-0351-1.
- Cherryholmes GA, Stanton SE, Disis ML. Current methods of epitope identification for cancer vaccine design. Vaccine. 2015;33(51):7408–14. doi:10.1016/j.vaccine.2015.06.116.