Generative AI-enabled neoantigen vaccine engineering: from tumour antigen discovery to personalized construct design and translational validation

A review in Biotechnology Advances describes how generative models may support the design of therapeutic neoantigen vaccines, from tumour antigen discovery to multi-epitope construct engineering. The authors contrast this with the discriminative models used so far, which only score and rank pre-existing mutant peptides by HLA binding, whereas a generative model explores and iteratively optimises sequence space under explicit constraints. The paper states explicitly that generative models should act as components of a wider workflow backed by experimental validation — including immunopeptidomics-guided calibration — not as a replacement for it. This is a review article and reports no clinical results of its own; the authors name limited validated datasets and regulatory interpretability demands as current barriers.