In-depth analysis · September 2026
Individualised cancer vaccines: a landscape reaching an inflection point
The first positive Phase 3 changes the nature of the bet. Attention is shifting towards indications, product architecture and the ability to manufacture a unique medicine for each patient.
All therapies mentioned are experimental; none is currently approved.
01
What changed in five weeks
On 19 August 2026, the field crossed its first proof threshold with a positive Phase 3 involving more than one thousand patients in adjuvant melanoma.
Nine days later, a Phase 2 colorectal study was stopped for futility with an overall-survival imbalance. On 2 September, Transgene reported 100% three-year disease-free survival among only sixteen treated patients.
The therapeutic class is gaining credibility, without definitive overall-survival evidence. Differentiation is moving towards the product, process and n-of-1 quality control.
02
The target and the principle
A neoantigen is a peptide created by a somatic mutation, presented by tumour-cell HLA and absent from the normal repertoire. In principle, this specificity limits the risk of targeting healthy tissue.
Because the mutanome is almost unique to each patient, the medicine is made as a single unit. Upstream steps converge: tumour/normal sequencing, transcriptomics, variant calling, HLA typing, presentation prediction, then ranking by expression and clonality.
Nearly all positive regimens combine a checkpoint inhibitor. The adjuvant setting, with minimal residual disease and enough manufacturing time, is currently the most favourable context.
neoantigens encoded across platforms, versus a published median of roughly three actually recognised in responding patients
03
Three architectures, three pharmaceutical objects
Neoantigen selection has largely converged. The difference lies in how the list is carried and how the lot is released.
| Architecture | Final object | Potential strengths | Challenges |
|---|---|---|---|
| Concatenated RNA · Moderna, BioNTech | One or two long RNAs: up to 34 epitopes for V940 and 20 for BioNTech | One synthesis, one formulation, one release | Limited target count, fixed ratios, artificial junctions and more difficult integrity as length increases |
| Short-RNA pool · OncoRNA | n monotope RNAs co-produced as one pool and one finished product | No junctions; short transcripts; potentially more targets; RNAseq pool control | Ratios must be controlled in the process and the regulatory strategy for a variable mixture validated |
| Viral vector · Transgene, Nouscom | One cassette carried by a viral particle | Intrinsic adjuvancy and multi-epitope presentation | Cell-culture production, anti-vector immunity and a cassette fixed at cloning |
Target count is becoming a tumour-diversity issue
Tumours evolve and contain multiple clones. A construct limited to a few dozen neoantigens cannot represent their full complexity. V940 announces up to 34 targets; BioNTech distributes up to 20 across two RNAs. This supports a compact product, but leaves tumour diversity, clonality and evolution incompletely covered.
The OncoRNA prototype brought together 121 distinct RNAs. Its pool architecture broadens target representation without lengthening one molecule. The project’s preclinical comparisons also showed a superior immune response versus the concatenated-RNA model.
04
Where do the main players stand?
Public status of programmes reviewed as of 28 September 2026.
| Company | Platform | Setting | Stage | Latest public signal |
|---|---|---|---|---|
| Moderna + Merck | Concatenated RNA ≤34, LNP, IM | Adjuvant melanoma | Phase 3 | RFS and DMFS met, 19 Aug 2026 |
| BioNTech + Genentech | 2 RNAs ≤20, lipoplex, IV | Adjuvant pancreatic cancer | Phase 2 ongoing | Phase 1: median RFS not reached among responders |
| BioNTech + Genentech | Same platform | ctDNA+ colorectal cancer | Phase 2 stopped | Futility and survival imbalance, 28 Aug 2026 |
| Transgene + NEC | MVA ≤30 | Adjuvant HPV-negative head and neck cancer | Phase 1/2 | 16/16 recurrence-free at 3 years vs 13/16 |
| Transgene | MVA, cell-line process | NSCLC | Phase 1 | Entered clinic, June 2026 |
| Nouscom | Adenoviral ≤60 | Melanoma, PD-L1 >50% NSCLC | Phase 1b | 90% manufacturing success claimed |
| Evaxion + MSD | Synthetic peptides | Advanced melanoma | Phase 2 | 75% ORR (12/16), ESMO 2025 |
| Geneos | Plasmid DNA + IL-12 | Hepatocellular carcinoma | Phase 1/2 | Nature Medicine publication, 2024 |
| Likang Life Sciences | RNA on autologous dendritic cells | Solid tumours | China clinical + FDA IND | FDA IND, February 2025 |
| Nykode · VB10.NEO | Plasmid DNA | Solid tumours | Phase 1/2 | Programme deprioritised since late 2024 |
Moderna and Merck
At five years, KEYNOTE-942 reported 68.8% recurrence-free patients with the combination versus 49.1% with pembrolizumab alone, and descriptive overall survival of 92.2% versus 71.3%. Phase 3 met RFS and DMFS, but effect size remains unpublished.
BioNTech and Genentech
In pancreatic cancer, eight responders among sixteen vaccinated patients had not reached median recurrence-free survival, versus 13.4 months among non-responders. IMcode003 must test that signal. The colorectal stop records a failed monotherapy in an MSS tumour, with a survival signal that requires serious scrutiny.
Transgene and NEC
The head-and-neck signal — sixteen of sixteen recurrence-free at three years — is striking but rests on three events and a small cohort. Vaccine responses were detected in 73.3%, with a median of three recognised neoantigens. The randomised two-year readout is expected in early 2028.
05
The 2026 inflection point
Between January and September — particularly across five weeks — the sector produced an unusual concentration of decisive signals.
Five-year KEYNOTE-942: HR 0.51
TG4050 Phase 2 fully recruited
First positive Phase 3 in melanoma
BioNTech colorectal programme stopped
TG4050: 100% three-year DFS in 16 patients
Transgene expands and industrialises myvac
- Class risk is falling: indications, combinations and production cost become the central questions.
- Clinical risk remains: tumour type, mutational burden and checkpoint combination still determine outcomes.
- Attention is moving to industrialisation: process design is becoming a core part of value.
06
Reading the evidence rigorously
Surrogate endpoints
Phase 3 met recurrence-free and distant-metastasis-free survival, not yet overall survival in a trial powered for that endpoint.
A favourable setting
Resected melanoma combines high mutational burden, immunosensitivity, minimal residual disease and enough manufacturing time. Extrapolation to cold tumours is not established.
Attributing benefit
Positive regimens include a checkpoint inhibitor. The vaccine’s independent contribution has not been isolated.
Three systemic biases
Responder comparisons confound treatment with immune fitness; small cohorts magnify each event; excluding manufacturing failures or early relapses can select better-prognosis patients.
Open biological questions
It remains unclear whether encoding more increases the chance of selecting the right targets or dilutes responses, how CD4 and CD8 responses divide, and how HLA loss, antigen loss or subclonal mutations should be managed.
Next hard readouts
INTerpath-004 in renal cancer (April 2027), IMcode003 in pancreatic cancer, Transgene’s randomised Phase 2 in head and neck cancer (Q1 2028), and INTerpath-012 in advanced melanoma (July 2028).
07
The industrial bottleneck: from scale-up to scale-out
Time
The sector target is six to eight weeks from biopsy to first injection. That timeline directly determines accessible indications.
Cost
Sipuleucel-T launched at $93,000 with production costs near 77% of price, showing that approval alone does not make a personalised product viable.
Lot release
Every lot is a new molecular entity. Regulators therefore assess a process and a generic, rapid and robust analytical method.
08 · OncoRNA
OncoRNA’s positioning
OncoRNA, an OncoDNA spin-out, approaches the problem through process design: co-producing short monotope RNAs in one operation and releasing them as one pool.
- 1Tumour/normal sequencing, variants, HLA, prediction and ranking
- 2Generate one template for each selected epitope
- 3Transcribe all templates in vitro in one reaction
- 4Formulate and release the lot as a pool
What the architecture could change
No artificial junctions
Produced sequences correspond to selected targets, without linkers that could create unwanted sequences.
Short transcripts
Adding a target adds another short species without lengthening an already complex molecule. This could represent tumour diversity more broadly.
A controllable ratio
Stoichiometry becomes a process parameter to measure and optimise rather than a fixed ratio within a concatemer.
Sequencing-based pool control
Pool RNAseq is intended to confirm in one measurement the nucleotide identity of each species, all expected sequences, no unexpected species and their relative abundance.
The method remains the same from patient to patient; only the reference file changes. This generic quality is precisely what an n-of-1 medicine needs.
Four regulatory questions to resolve
- Define acceptance limits for relative pool composition.
- Control template stoichiometry upstream because a species cannot be removed after co-transcription.
- Validate RNAseq and its bioinformatics pipeline as a GMP release method.
- Define a relevant potency assay, still an open question across the field.
At the preclinical stage, OncoRNA combines broader target coverage, a superior immune response versus the concatenated-RNA model in the comparisons performed, and a process designed to improve simplicity, time and cost.
09
Why now?
Merck and Moderna have enrolled roughly 4,660 patients around one individualised product across five organs, with readouts through 2034. Their success helps validate the principle at leaders’ expense; OncoRNA’s opportunity concerns the still-open industrial execution.
The regulatory standard will be set
The first filing will force regulators to define n-of-1 process, release and comparability expectations. Engaging before those expectations harden may matter.
Consolidation is under way
Gritstone bio was liquidated, Achilles and Genoea stopped, Neon was acquired and BioNTech bought CureVac. Credible independent platforms are becoming scarcer.
Industry is still finding its model
Moderna is building a dedicated facility; Transgene redesigned its process for transferability and automation. The industrial position remains open.
Questions from an informed investor
| Objection | Honest answer |
|---|---|
| The giants are ten years ahead | Clinically, yes. In process and release, they are still redesigning their own systems. |
| The market is still tiny | True: no individualised cancer vaccine is approved. A first registration is precisely what could change that. |
| Regulation is not ready for a pool | True, and that is the work under way. No architecture in the class has a registered product yet. |
| What is the advantage over a concatemer? | The project’s preclinical comparisons showed a superior immune response for the pool. Its architecture can also broaden target coverage without lengthening one molecule. |
Five deliverables that would turn the case into a dossier
- Pool-composition reproducibility across representative lots
- Qualification of RNAseq and its bioinformatics pipeline for release
- Measured rather than estimated biopsy-to-lot time
- Preclinical immunogenicity comparison between pool and concatemer
- Formal regulatory scientific advice on the control strategy
10
Review sources
Pages consulted as of 28 September 2026.
Moderna / Merck
- INTerpath-001 RFS and DMFS result, 19 August 2026
- Five-year KEYNOTE-942 data and ASCO 2026 presentation
- INTerpath background dossier and analyses of the Phase 3 result
BioNTech / Genentech
- Autogene cevumeran colorectal Phase 2 stop, 28 August 2026
- Nature: persistence of CD8+ cells in pancreatic cancer
- Phase 1 data and long-term follow-up
- Completion of CureVac acquisition, 2025
Transgene / NEC and others
- TG4050 three-year recurrence-free survival, Nature Communications, 2026
- myvac platform extension, 24 September 2026
- Nouscom, Evaxion, Geneos and Chinese personalised-RNA pipelines
Industrialisation and market
- Analysis of manufacturing barriers for personalised RNA vaccines
- Personalised vaccine market estimates and forecasts
Two sources were inaccessible during the review: the ESMO 2025 IMcode001 abstract and a 2026 Asia-Pacific Journal of Clinical Oncology pipeline review.
In summary
A class being validated, a differentiated architecture
Leaders’ success can expand the market for every credible participant. OncoRNA differentiates through more targets, a superior immune response versus the concatenated-RNA model in the project’s preclinical comparisons, and a process designed to be simpler, faster and less costly.