Who’s building neoantigen cancer vaccines (2026)

Neoantigen cancer vaccines — immunotherapies that teach a patient's T cells to recognize the mutated proteins specific to their tumor — have moved from a promising idea to a field with multiple Phase 2 and Phase 3 programs and a string of near-term catalysts. The momentum is real: durable randomized data in adjuvant melanoma, immune-response signals across pancreatic, liver, and head-and-neck cancers, and an off-the-shelf approach now validated in a peer-reviewed prevention setting. The caution is equally real: no product is approved, several pivotal endpoints are still immature, and at least one high-profile program has hit a futility boundary.

The field splits along two axes worth keeping straight. First, personalized (individualized) vaccines — sequenced and manufactured per patient — versus off-the-shelf vaccines that target neoantigens shared across many tumors. Second, the delivery platform: mRNA (Moderna, BioNTech), viral vector (Transgene, Nouscom), peptide (Evaxion), or DNA (Geneos). Almost every player now leans on an AI or computational target-selection layer; the differentiation increasingly lives in how well that layer picks neoantigens that actually provoke a tumor-specific response. This landscape maps who is building what, how mature each readout is, and what to watch.

Clinical timelines in this field shift constantly, and the table above is a point-in-time snapshot. For a continuously refreshed view — every neoantigen and personalized-cancer-vaccine trial with its current phase, recruitment status, indication, and last-update date — see the live tracker, rebuilt daily from ClinicalTrials.gov:

• Neoantigen vaccine clinical trials tracker — live, auto-updating; the headline programs above are pinned to the top.

Moderna and Merck's intismeran autogene (formerly mRNA-4157 / V940) is the field's frontrunner. The randomized Phase 2b KEYNOTE-942 trial enrolled 157 patients with resected high-risk stage III/IV melanoma and tested the personalized mRNA vaccine on top of KEYTRUDA (pembrolizumab) versus pembrolizumab alone. At roughly three years' median follow-up the combination reduced the risk of recurrence or death by 49% and the risk of distant metastasis or death by 62% versus pembrolizumab alone; the companies have since reported sustained recurrence-free survival benefit at five years (verify the latest cut against the primary source). These are clinical-efficacy and event-driven survival-proxy endpoints (RFS, DMFS) in a mid-size randomized trial — stronger than immune-response data alone, but still Phase 2b. The pivotal test is Phase 3: INTerpath-001 in adjuvant melanoma is fully enrolled with data potentially in 2026 (verify), while INTerpath-002 and INTerpath-009 extend the bet into resected NSCLC and are still enrolling, alongside additional tumor types.

BioNTech's autogene cevumeran (BNT122), partnered with Genentech/Roche, is the other personalized-mRNA leader. Its standout signal is in pancreatic cancer (PDAC): in a Phase 1 study, 8 of 16 patients mounted vaccine-induced T-cell responses, and those responders showed longer median recurrence-free survival, with immune responses persisting up to three years. That signal is now being tested in the randomized Phase 2 IMCODE003 trial (~260 resected PDAC patients, vaccine plus atezolizumab and mFOLFIRINOX versus chemotherapy alone). Important caveat: the Phase 1 responder/non-responder split is a small, non-randomized subgroup observation, not proof of efficacy — IMCODE003 exists precisely to test whether it is reproducible.

Evaxion (NASDAQ: EVAX) leads with EVX-01, a personalized peptide vaccine designed by its AI-Immunology platform, in metastatic melanoma on top of pembrolizumab. The company reported that 86% of AI-selected vaccine targets triggered a tumor-specific immune response, and earlier Phase 2 data showed a 75% objective response rate (12 of 16 patients, 4 complete responses), with two-year follow-up indicating most responders remained without relapse. Read these carefully: the ORR comes from a small, single-arm cohort with no internal control, so it cannot be cleanly attributed to the vaccine versus the checkpoint-inhibitor backbone. The platform-precision metric is an immune-response endpoint, not a clinical one. Three-year clinical-efficacy data are expected in H2 2026 (verify).

Transgene, with NEC providing AI-driven neoantigen prediction, is developing TG4050, a personalized vaccine built on a viral-vector (modified vaccinia Ankara) platform, in adjuvant HPV-negative head-and-neck cancer. The Phase 1 portion met its endpoints and induced durable neoantigen-specific immune responses sustained up to two years, with published data supporting a relapse-prevention rationale. Randomization in the Phase 2 portion is complete, and the first Phase 2 immunogenicity data are expected in H2 2026 (verify), with the primary endpoint readout later. As with the others, the early-stage strength here is immunogenicity and small-sample relapse signals, not yet randomized efficacy.

Nouscom takes the opposite tack from the personalized leaders. NOUS-209 is an off-the-shelf viral-vector vaccine (gorilla adenovirus prime, MVA boost) encoding ~209 shared frameshift-peptide neoantigens that recur across mismatch-repair-deficient (dMMR) and microsatellite-instable (MSI) tumors. Because the antigens are shared, one fixed product can be manufactured in advance — sidestepping the per-patient sequencing and manufacturing logistics that make personalized vaccines expensive and slow. Results from a Phase 1b/2 trial in Lynch-syndrome carriers were published in Nature Medicine (2026), reporting that the vaccine was safe and elicited potent, durable neoantigen-specific T-cell responses. The framing is notable: this targets cancer interception/prevention in a high-risk hereditary population (Lynch carriers face up to ~80% lifetime colorectal-cancer risk), a distinct and ambitious use case versus treating established disease. The published data are immunogenicity and safety; prevention efficacy would require far longer follow-up.

The single most important cautionary datapoint is BioNTech/Genentech's autogene cevumeran in adjuvant colorectal cancer (the Phase 2 BNT122-01 trial in ctDNA-positive, resected stage II-high-risk/III CRC). At a first interim analysis the trial crossed a boundary for futility. The companies kept it blinded and ongoing because the data were judged not yet mature enough for reliable efficacy conclusions, but the final analysis has slipped from 2026 to 2027 as events accrued more slowly than projected. Read straight, that is a negative interim signal in CRC — not a positive one. The context that matters: most CRC is microsatellite-stable (MSS), an immunologically 'cold' setting with low mutational burden and few neoantigens, where checkpoint inhibitors have largely failed and a neoantigen vaccine has less raw material to work with. The same platform trending positive in PDAC and strongly positive in melanoma does not transfer automatically to MSS-CRC; biology, not execution, is the likely constraint. Don't bury the setback, and don't cheerlead around it.

Beyond that, three field-wide risks: endpoint maturity — much of the bullish data is immune-response or small-sample ORR/RFS, not mature randomized overall survival, and Phase 2 signals have historically not always survived Phase 3; manufacturing and logistics — personalized vaccines require per-patient sequencing, prediction, and bespoke production on a clinically useful timeline; and commercial viability — Gritstone bio, an early neoantigen pioneer, filed for Chapter 11 in October 2024 and sold substantially all its assets, a reminder that scientific promise and capital survival are different questions. The adjacent IO Biotech Phase 3 (an off-the-shelf IDO/PD-L1 peptide vaccine, not a classical neoantigen product) narrowly missed statistical significance on PFS in first-line melanoma (p=0.056) in August 2025 — a useful reminder of how thin the margin between 'positive' and 'missed' can be even with a real clinical effect.

Net: the field has graduated from 'can a neoantigen vaccine induce an immune response' (largely yes) to 'does that translate into durable clinical benefit in randomized trials' (being tested now). 2026-2027 is the proving window — INTerpath in melanoma and NSCLC, IMCODE003 in PDAC, the BNT122-01 final analysis, and H2-2026 mid-cap readouts from Evaxion and Transgene.

Last reviewed 2026-05-30. This is a fast-moving field and clinical timelines shift frequently; trial stages and especially data-readout dates are provisional. Verify any stage, result, or date against the primary source (company press releases, ClinicalTrials.gov, and peer-reviewed publications) before relying on it for decisions. Items marked '(verify)' are forward-looking guidance we could not confirm as final.

• Moderna/Merck — 3-year mRNA-4157 (V940) melanoma data
• BioNTech — 3-year Phase 1 follow-up, autogene cevumeran in PDAC
• ApexOnco — BioNTech/Roche CRC futility-boundary analysis
• Transgene — Phase 2 randomization complete in head & neck (TG4050)
• Evaxion — 86% AI-Immunology target precision, EVX-01 Phase 2
• Nouscom — NOUS-209 Phase 1b/2 in Nature Medicine