Leveraging Artificial Intelligence for Neoantigen Prediction
Jing Zeng, Zhengjun Lin, Xianghong Zhang, Tao Zheng, Haodong Xu, Tang Liu
Abstract
Neoantigens represent a class of antigens within tumor microenvironments that arise from diverse somatic mutations and aberrations specific to tumorigenesis, holding substantial promise for advancing tumor immunotherapy. However, only a subset of neoantigens effectively elicits antitumor immune responses, and the specific neoantigens recognized by individual T-cell receptors (TCR) remain incompletely characterized. Therefore, substantial research has focused on screening immunogenic neoantigens, mainly through their major histocompatibility complex (MHC) presentation and TCR recognition specificity. Given the resource intensiveness and inefficiency of experimental validation, predictive models based on artificial intelligence (AI) have gradually become mainstream methods to discover immunogenic neoantigens. In this article, we provide a comprehensive summary of current AI methodologies for predicting neoantigens, with a particular focus on their capability to model peptide-MHC (pMHC) and pMHC-TCR binding. Furthermore, a thorough benchmarking analysis was conducted to assess the performance of antigen presentation predictors for scoring the immunogenicity of neoantigens. AI models have potential applications in the treatment of clinical diseases although several limitations must first be overcome to realize their full potential. Anticipated advancements in data accessibility, algorithmic refinement, platform enhancement, and comprehensive validation of immune processes are poised to enhance the precision and utility of neoantigen prediction methodologies. This article is part of a special series: Driving Cancer Discoveries with Computational Research, Data Science, and Machine Learning/AI.