Introduction: Messenger RNA (mRNA) vaccines represent one of the most significant innovations in contemporary biomedical science. Initially recognized for their effectiveness in preventing infectious diseases such as COVID-19, these vaccines are now being explored as powerful tools in cancer immunotherapy and other therapeutic applications. Their advantages include rapid design, adaptability, and induction of strong immune responses. However, despite their promise, clinical translation faces challenges due to genetic heterogeneity, tumor immune evasion, and the vast complexity of multi-omics datasets. Predicting patient responses to mRNA vaccines and immunotherapies remains a critical unmet need. Artificial intelligence (AI), through advanced machine learning and deep learning methods, offers new opportunities to integrate heterogeneous biological, clinical, and imaging data to improve accuracy in patient-specific response prediction.
Methods: This review synthesizes findings from studies published between 2019 and 2025 on AI-based prediction of responses to mRNA vaccines and immunotherapy. We examined multiple layers of data including genomic (whole-genome sequencing, whole-exome sequencing, SNP panels), transcriptomic (bulk and single-cell RNA-seq, spatial transcriptomics), proteomic, microbiome (16S rRNA sequencing, metagenomics, metabolomics), imaging (CT, MRI, PET, histopathology, radiomics), and clinical-demographic datasets. We evaluated classical machine learning algorithms such as support vector machines (SVM), random forests, and gradient boosting, alongside deep learning architectures including convolutional neural networks (CNNs), recurrent neural networks (RNNs), transformers, and biologically disentangled variational autoencoders (BDVAE). We also reviewed mechanistic immune modeling, explainable AI (XAI), and federated learning approaches for privacy-preserving collaborations.
Results: AI and machine learning methods have shown remarkable accuracy in predicting immunotherapy outcomes. In melanoma, SVM models achieved concordance indices above 0.76, while random forest classifiers reported up to 0.89 in treatment response prediction. CNN-based radiomics models analyzing CT data reached AUC values of 0.94 in non-small cell lung cancer. Transformer-based architectures integrating gene expression and DNA methylation profiles achieved similarly high predictive performance in pan-cancer cohorts. Multi-omics integration models such as BDVAE not only improved accuracy (AUC = 0.94) but also uncovered critical resistance mechanisms including metabolic remodeling and immune suppression. Clinical trials of personalized mRNA vaccines, such as autogene cevumeran in pancreatic cancer, demonstrated that a subset of patients mounted strong immune responses with significantly prolonged survival, aligning with AI-predicted biomarkers of responsiveness. Furthermore, microbiome analyses revealed that microbial diversity and metabolites like short-chain fatty acids significantly influence treatment efficacy, providing novel predictive features for AI models. Federated learning has emerged as a feasible solution to address privacy and data-sharing limitations in multi-institutional research, while explainable AI techniques such as SHAP and LIME have improved model interpretability and clinical trust.
Conclusion: AI-driven prediction of patient responses to mRNA vaccines and immunotherapy demonstrates significant promise in advancing precision medicine. Integrating genomic, transcriptomic, proteomic, microbiome, clinical, and imaging data through advanced ML/DL models provides a holistic framework for individualized treatment strategies. While challenges such as class imbalance, data heterogeneity, and ethical concerns remain, innovations including multi-omics fusion, mechanistic modeling, federated learning, and explainable AI offer robust solutions. Together, these strategies highlight a transformative pathway toward secure, interpretable, and equitable application of mRNA vaccines and immunotherapies in diverse patient populations.
Keywords: mRNA vaccines; immunotherapy; artificial intelligence; multi-omics; precision medicine