News Release

Cancer vaccines built from each patient's own tumor mutations are entering their clinical era

Nanjing Drum Tower Hospital team reviews how AI-driven neoantigen prediction and six delivery platforms — from mRNA to engineered bacteria — are shaping personalized immunotherapy

Peer-Reviewed Publication

Innovation Press

Current landscape and future directions of neoantigen vaccines: A new era of personalized cancer immunotherapy

image: 

  • Neoantigens are tumor-specific and highly immunogenic targets for personalized cancer immunotherapy.
  • Artificial intelligence (AI) improves neoantigen prediction, yet challenges remain in accuracy.
  • Diverse vaccine platforms show promise in clinical trials.
  • Combination therapies are crucial for enhancing antitumor efficacy of neoantigen vaccines.
view more 

Credit: Wang X., Su S., Tan S., et al., (2026). Current landscape and future directions of neoantigen vaccines: A new era of personalized cancer immunotherapy. The Innovation 7(9), 101357.

Most cancer therapies hurt healthy tissue along the way. A different idea has been quietly gathering force: vaccines built from each patient's own tumor mutations, designed to teach the immune system to strike the cancer and spare everything else. A review in The Innovation surveys where this field stands — and what it must still solve.

The targets are neoantigens: abnormal peptides that cancer cells alone produce, arising from somatic mutations such as single-nucleotide variants, insertions and deletions, and gene fusions. Because they do not exist in normal tissue, they escape the immune tolerance that blunts vaccines aimed at self antigens, and can drive potent, tumor-specific T cell responses. The catalog keeps expanding — recent work reveals neoantigens arising from noncoding-region translation, aberrant RNA splicing, and protein modifications, widening the pool of druggable targets.

The hard part is choosing. A tumor can carry thousands of mutations, but only a fraction yield peptides the immune system can actually see and attack. Prediction has therefore moved from simple MHC-binding scores toward models that integrate T-cell receptor recognition and multi-omics features; newer deep-learning tools such as FL-RNC-seq and deepAntigen are raising the hit rate and cutting false positives. Accuracy, however, remains imperfect — the field's most-cited open problem, the review notes.

Delivery is the second engine. The review compares six platforms, each with tradeoffs: peptide and protein vaccines are safe and easy to personalize but weakly immunogenic; mRNA vaccines wrapped in lipid nanoparticles are fast to manufacture and scale, yet degrade quickly; dendritic-cell vaccines present antigen efficiently at high cost; viral vectors fire strong T cell responses without adjuvants and have shown survival benefit in colorectal cancer trials; engineered bacteria such as Escherichia coli Nissle 1917 can home to oxygen-starved tumor cores, though their clinical safety record is still thin; and self-assembling nanoparticles passively target lymph nodes while carrying many epitopes at once.

Clinical experience supplies the third lesson: successful regimens combine multi-epitope designs that cover tumor heterogeneity with immune checkpoint inhibitors, and pair vaccines with radiotherapy or metabolic modulators to soften the tumor microenvironment's physical barriers and suppressive cytokines. The remaining bottleneck is the "cold tumor" — cancers that attract few immune cells and respond poorly.

The authors frame the moment as a golden age, moving neoantigen vaccines from proof of concept toward clinical breakthroughs. The path forward runs through the same three engines: sharper AI prediction, better delivery, and smarter combinations. For patients, the promise is concrete — a therapy matched to the unique mutation set of their own tumor, with the potential for durable control rather than a blunt assault. The challenges are equally concrete: prediction accuracy that still falls short, platforms that remain hard to manufacture at scale, clinical evidence that must yet mature across cancer types, and the sheer cost of a personalized product — long-peptide synthesis, with its purity and expense problems, is a telling example.


Disclaimer: AAAS and EurekAlert! are not responsible for the accuracy of news releases posted to EurekAlert! by contributing institutions or for the use of any information through the EurekAlert system.