Artificial intelligence and machine learning in the development of vaccines and immunotherapeutics-yesterday, today, and tomorrow

Abstract

The development of vaccines and immunotherapies against infectious diseases and cancers has been one of the significant achievements of medical science in the last century. Subunit vaccines offer key advantages over whole-inactivated or attenuated-pathogen-based vaccines, as they elicit more specific B-and T-cell responses with improved safety, immunogenicity, and protective efficacy. However, developing subunit vaccines is often cost-and time-consuming. In the past, the development of vaccines and immunotherapeutics relied heavily on trial-and-error experimentation, as well as extensive and costly in vivo testing, which typically required years of pre-clinical and clinical trials. Today, artificial intelligence (AI) and deep learning (DL) are actively transforming vaccine and immunotherapeutic research by (i) offering predictive frameworks that support rapid, data-driven decision-making, (ii) integrating computational models, systems vaccinology, and multi-omics data (iii) helping to better phenotype, differentiate, and classify patients diseases and cancers; (iv), integrating host characteristics for tailored vaccines and immunotherapeutics; (v) refining the selection of B-and T-cell antigen/epitope targets to enhance efficacy and durability of immune protection; and (vi) enabling a deeper understanding of immune regulation, immune evasion, and regulatory pathways. Artificial intelligence and DL are pushing the boundaries toward (i) the potential replacement of animal preclinical testing of vaccines and immunotherapeutics with computational-based models, as recently proposed by the United States NIH and FDA, and (ii) improving clinical trials by enabling real-time modeling for immune-bridging, predicting patients' immune responses, safety, and protective efficacy to vaccines and immunotherapeutics. In this review, we describe the past and current applications of AI and DL as time-and resource-efficient strategies and discuss future challenges in implementing AI and DL as new transformative fields that may facilitate the rapid development of precision and personalized vaccines and immunotherapeutics for infectious diseases and cancers.

Keywords: artificial intelligence (AI); cancer; immunotherapeutics; infectious diseases; machine learning (ML); vaccine development.

Figure 1

AI-driven framework optimizes b-and t-cell epitope prediction, classification, and multi-epitope vaccine design. (A) Epitope prediction model (transformer-based) for identifying immunogenic B-cell and T-cell epitopes, (B) deep learning epitope classification using (CNN-based) biomarker extraction and feature selection, (C) a multi-task autoencoder prioritizes immunogenic and protective epitopes, and (D) AI-powered vaccine formulation integrating top-ranked asymptomatic B-and T-cell epitopes into a candidate vaccine using Generative Adversarial Networks (GANs). GANs refine and generate four multi-epitope-based next-generation coronavirus vaccine candidates.

Figure 2

AI-driven workflow for immunology research. This illustration presents a structured framework outlining the role of Artificial Intelligence (AI) in immunology research, demonstrating how AI-driven approaches enhance immunological analysis, predictive modeling, and therapeutic development. The process begins with the identification of key immunological challenges, such as disease prediction and vaccine development, followed by the collection and preprocessing of immunological datasets, including gene expression profiles and protein interactions. Feature selection then plays a crucial role in identifying relevant biomarkers and immune-related variables to improve model accuracy. Once the data is prepared, machine learning and deep learning models are built and trained on immune system datasets to recognize patterns and predict immune responses. The trained models undergo evaluation and optimization using performance metrics like ROC-AUC and precision-recall curves to ensure accuracy and reliability. Ultimately, these AI-driven models are applied in real-world immunology research, enabling advancements in personalized medicine, immunotherapy, and vaccine optimization. By integrating AI techniques into immunology, researchers can gain deeper insights into immune system behavior, develop more effective therapeutic interventions, and refine disease treatment strategies.

Figure 3

AI-driven mathematical model of CD8⁺ T cell response to viral infections: T (Blue) – CD8⁺ T cells that recognize and eliminate infected cells; I (Red) – Virus-infected cells that replicate and produce viral particles; E (Green) – Effector immune cells (e.g., activated CD8⁺ T cells); V – Free virus particles that infect host cells; βT – Rate at which virus infects host cells and generates infected cells (I); βTV – Interaction coefficient representing how CD8⁺ T cells (T) are stimulated by viral load (V); p – Rate of virus production by infected cells (I); κIE – Killing rate of infected cells (I) by effector T cells (E); ρI – Stimulation coefficient, representing the activation of effector immune cells (E) by infected cells (I); Cv – Clearance rate of virus (V) due to immune response.

Figure 4

Growth of AI-driven research in immunology (2015–2024). (A) Conceptual representation of the hierarchical relationship among Artificial Intelligence (AI), Machine Learning (ML), and Deep Learning (DL) in the context of immunological applications. AI serves as the broadest category, encompassing ML techniques that identify patterns in immune-related datasets. DL, a specialized subfield of ML, is used to model complex immune responses and uncover pathways critical for vaccine efficacy and immune regulation. (B) The integration of Artificial Intelligence (AI) in immunology has undergone significant expansion over the past decade. The number of publications in this field has grown from just over 100 articles per year in 2015 to over 1,500 in 2024, highlighting the increasing role of AI in immunological research and its applications in vaccine design, immune response modeling, and precision medicine. This diagram was generated based on keyword searches such as “AI in immunology,” “machine learning in immunology,” and “deep learning in immune research” from the PubMed database. The search encompassed peer-reviewed articles, systematic reviews, and conference proceedings related to AI-driven immunological studies. The data reflects the rising trend in AI applications, including areas such as predictive immune modeling, AI-assisted diagnostics, and computational vaccine development. The exponential rise in publications post-2020 aligns with breakthroughs in deep learning, large-scale immunological datasets, and AI-driven drug discovery, indicating a paradigm shift in how computational tools are being utilized in immunology. The projected increase in AI-based studies suggests continued advancements in immune system modeling, personalized immunotherapy, and AI-enhanced vaccine development strategies.

Figure 5

AI-driven insights into tumor-immune interactions and memory T cell differentiation. (A) AI’s Role in Tumor-Immune Interactions and Checkpoint Blockade: This panel illustrates how Artificial Intelligence (AI) supports understanding and optimization of tumor-immune interactions, particularly in regulating immune checkpoint inhibitors. Tumor cells suppress T cell activity through checkpoint pathways such as PD-1/PD-L1 and CTLA-4. AI models predict optimal checkpoint blockade strategies by identifying patient-specific responses, discovering biomarkers, and enhancing the efficacy of combination immunotherapies. Integration of multi-omics data (genomics, transcriptomics, proteomics) via AI enables personalized immunotherapy with improved efficacy and reduced toxicity. (B) AI Insights into Memory T-Cell Differentiation: This panel illustrates the role of AI in guiding the differentiation of memory T cells. Following antigen exposure, naïve T cells activate and differentiate into effector and memory T cells. AI enhances this process by predicting activation based on TCR signaling, analyzing differentiation markers, and optimizing memory formation. AI models trained on single-cell RNA sequencing and epigenetic data uncover pathways critical for long-term immune protection, informing vaccine development and cancer immunotherapies. (C) Interaction Between Dendritic Cells and T Cells in Antigen Presentation. This illustration depicts the crucial role of dendritic cells (DCs) in bridging innate and adaptive immunity through antigen presentation. The dendritic cell presents antigens to CD4⁺ T cells via MHC class II molecules, which are recognized by the T cell receptor (TCR) in conjunction with the CD4 co-receptor, facilitating the activation of helper T cells. Simultaneously, CD8⁺ T cells recognize antigens presented on MHC class I molecules, with the TCR engaging the complex alongside the CD8 co-receptor, resulting in the activation of cytotoxic T cells. This dual interaction is essential for initiating and coordinating immune responses, enabling helper T cells to support other immune cells and cytotoxic T cells to eliminate infected or malignant cells.

Figure 6

Antigen/epitope prediction model workflow. This figure illustrates the structured workflow of the transformer-based deep learning model used for antigen and epitope prediction. The pipeline begins with input data, consisting of viral protein sequences and epitope data, which undergo preprocessing to extract relevant features. The dataset is then split into training (80%) and testing (20%) subsets to ensure proper model generalization. The transformer-based model is trained and fine-tuned using deep learning techniques to improve prediction accuracy. Finally, the model undergoes performance evaluation using metrics such as Accuracy, Recall, and F1-score, providing insight into the reliability of predicted immunogenicity scores. This structured approach ensures that the model accurately distinguishes between highly immunogenic and non-immunogenic epitopes, facilitating the selection of vaccine targets.

Figure 7

Training and validation performance of the antigen/epitope prediction model: (A) Loss and (B) Accuracy. This figure presents the training performance metrics of the antigen/epitope prediction model. The left graph displays the training and validation loss over 100 epochs, where both losses decrease progressively, indicating stable optimization. The validation loss stabilizes at 0.032, demonstrating strong generalization without overfitting. The right graph showcases training and validation accuracy, which steadily improved, reaching 0.993 training accuracy and 0.935 validation accuracy. These results confirm that the model effectively learns to recognize key features of immunogenic and non-immunogenic epitopes while maintaining high reliability across validation data. The minimal gap between training and validation curves highlights robust model performance, ensuring applicability for real-world antigen screening.

Figure 8

Artificial intelligence is revolutionizing the pre-clinical and clinical development of vaccines and immunotherapeutics. The traditional pathway (top panels) relies heavily on empirical antigen discovery, animal-based preclinical testing, conventional clinical trial designs, and retrospective regulatory evaluation, which can lead to ethical concerns, interspecies variability, and prolonged development timelines. In contrast, the AI-powered pathway (bottom panels) illustrates how emerging computational models can replace or reduce the need for animal preclinical testing, as endorsed by the United States FDA under the Modernization Act 2.0. AI-driven platforms leverage deep learning to predict protective antigens, simulate human immune responses in silico, and enable real-time immune-bridging strategies during clinical trials. These advancements accelerate candidate selection, optimize clinical trial design, predict protection based on immune markers, and enhance the speed and precision of regulatory approval and post-market surveillance.