Merck and Moderna's mRNA cancer vaccine uses a neoantigen algorithm to personalize treatment for each patient

Merck and Moderna's investigational personalized cancer vaccine, mRNA-4157, relies on a computational algorithm to identify tumor-specific neoantigens from a patient's own cancer sequencing data and select the best targets to encode in the mRNA construct. The approach is currently advancing in melanoma and other solid tumor settings following promising Phase 2 data showing improved recurrence-free survival when combined with pembrolizumab.

Why it matters

The mRNA-4157 vaccine works by sequencing a patient's tumor, running that data through a proprietary algorithm to rank candidate neoantigens, and then encoding the top selections into a bespoke mRNA construct. Each patient effectively receives a different drug. That is a genuinely novel manufacturing and clinical paradigm, and the logistics alone, from biopsy to infusion turnaround time, are a real operational challenge that does not get enough attention in the enthusiasm around the efficacy signal.

What deserves more scrutiny is the algorithm itself. Neoantigen prediction is not a solved problem. Models must account for mutation calling accuracy, HLA typing, peptide-MHC binding prediction, and antigen processing, each step introducing its own error rate. The clinical results from KEYNOTE-942 are encouraging, but we do not yet know how much of the benefit comes from the vaccine versus the pembrolizumab backbone, and we do not have visibility into how often the algorithm selects neoantigens that the immune system actually responds to. Transparency on those questions will matter enormously as this moves toward broader use.

What to watch

Phase 3 trial results and whether the neoantigen prediction accuracy holds across tumor types beyond melanoma, where mutational burden and immune context differ substantially.