Neoantigen × AI
Daily research signal
Landscape · 2026-06-04

Vaccine or T cell? Two ways to drug a neoantigen

ASCO 2026 showed neoantigens being targeted not just by vaccines but by engineered TCR-T cells — the first objective responses against the TP53 R175H neoantigen. Both read the same tumor and pick the same targets; they differ in how the T cell gets made. A map of when each approach wins.

For most of the last decade, "targeting a neoantigen" meant one thing: a vaccine. Read a patient's tumor, work out which of its mutations produce peptides the immune system can see, and put those peptides — or the mRNA that encodes them — into a shot that teaches the patient's own T cells to attack. ASCO 2026 made it clear there is a second route. An engineered T-cell therapy, NT-175, reported the first objective responses against the TP53 R175H neoantigen — an objective response rate of 47.6% in a first cohort of 21 patients — by manufacturing T cells that already carry a receptor for that target, rather than coaxing the patient's immune system into making them.

That is worth a moment's attention from anyone tracking personalized cancer vaccines, because the two approaches are siblings, not strangers. They share the entire hard part of the problem and differ only at the very end. This piece lays out where they overlap, where they diverge, and — the question that actually matters for a patient — when you would reach for one over the other.

A neoantigen vaccine and a neoantigen-reactive T-cell therapy both begin with the identical pipeline. Sequence the tumor and healthy tissue to find the mutations that are unique to the cancer. Type the patient's HLA alleles. Predict which mutated peptides will actually be processed and presented on those HLA molecules — the step where AI binding and presentation models do the heavy lifting — and then guess which of the presented peptides a T cell can productively recognize. That last step, immunogenicity, is the field's central unsolved problem regardless of what you build at the end of it.

Everything our daily brief tracks — epitope and HLA-binding prediction, immunopeptidomics, immunogenicity ranking, TCR–pMHC modeling — feeds both modalities equally. A better predictor of which neoantigen is real makes a better vaccine and a better TCR-T. The divergence comes only after the target is chosen: do you teach the patient's immune system to find it, or do you build the T cell yourself and infuse it?

A vaccine is active immunization. It presents the chosen neoantigens to the patient's immune system and relies on that system to raise its own T-cell response — polyclonal, often against many targets at once (intismeran encodes up to 34 neoantigens per patient), and capable of laying down long-lived memory. The cost is dependence on the patient: if the immune system is exhausted, old, or suppressed by the tumor microenvironment, the response may be weak, and it takes weeks to build.

A TCR-T is adoptive transfer. You identify a T-cell receptor that recognizes the neoantigen, engineer the patient's (or a donor's) T cells to express it, expand them to large numbers, and infuse them after lymphodepletion. The payoff is control: you deliver a defined dose of high-avidity T cells that recognize the target whether or not the patient could have raised them, and you can armor the cells — NT-175, for instance, knocks out the receptor for TGF-β so the cells resist a key tumor-suppression signal. The costs are a complex, per-patient manufacturing process, the toxicity of lymphodepletion, and — because most designs carry a single receptor against a single target — the risk that the tumor escapes by losing that one antigen or its presenting HLA.

Neoantigen vaccineNeoantigen-reactive TCR-T
MechanismActive immunization — patient raises the responseAdoptive transfer — response is manufactured and infused
TargetsUsually many neoantigens at once (polyclonal)Usually one receptor, one target (monoclonal)
Depends on patient immune fitnessYes — must mount a responseNo — T cells supplied ready-made
OnsetWeeks (priming + boosts)Days after infusion
ManufacturingPeptide/mRNA — relatively lightCell product — heavy, per patient
ConditioningNone / standard adjuvantLymphodepletion
Escape riskLower — multiple targetsHigher — single antigen / HLA loss
DurabilityImmune memoryPersistence of infused cells (variable)
Active vs adoptive delivery of a neoantigen-directed T-cell response. The upstream target-selection problem is shared; these are the downstream trade-offs.

There is a second distinction that cuts across the first, and ASCO 2026 illustrated both ends of it. A neoantigen can be private — unique to one patient's tumor, the basis of a personalized vaccine like intismeran or NeoVax that must be manufactured per person. Or it can be shared (public) — the same mutation recurring across many patients. The most common cancer mutations, in genes like TP53 and KRAS, are shared, which is what makes them attractive off-the-shelf targets.

NT-175 sits at the shared-neoantigen, cell-therapy corner: a single fixed receptor against TP53 R175H, usable for any HLA-A*02:01 patient whose tumor carries that mutation — no per-patient target discovery required, only patient selection. NOUS-209, in the vaccine column, is the shared-neoantigen vaccine equivalent: one product encoding frameshift neoantigens common to mismatch-repair-deficient tumors. Personalized vaccines occupy the opposite corner. Personalized neoantigen TCR-T — building a bespoke receptor against a private mutation — is the fourth corner, and the one still mostly in development; it did not have a marquee ASCO 2026 readout. Seeing three of the four corners produce clinical data in a single meeting is the real signal: neoantigens have become a target class, addressable by more than one kind of medicine.

Keep the maturity of the evidence in view, because it differs sharply across the corners. The vaccine data are the most advanced: intismeran has a five-year, randomized, controlled readout in melanoma (recurrence risk down 49%), and NeoVax has a survival signal in glioblastoma against propensity-matched controls. The neoantigen-TCR-T data are early — NT-175's 47.6% response rate is a first-in-human, single-arm cohort of 21 patients, and a second program against the KRAS G12V neoantigen (presented under abstract 2542) was flagged as one to watch rather than a mature result. Early single-arm response rates and multi-year randomized survival curves are not the same currency, and should not be compared as if they were.

What the cell-therapy data do establish is feasibility: it is possible to build a T cell against a shared hotspot neoantigen and see tumors shrink. For targets that vaccines and small molecules have both struggled with — mutant p53 chief among them — that is a genuinely new option, even at an early stage.

The settings sort the two approaches more than the biology does. Vaccines are best where there is time and little disease: the adjuvant setting after surgery, minimal-residual-disease interception, and outright prevention (NOUS-209 in Lynch syndrome). There the patient is relatively immunocompetent, the tumor burden is low, and the weeks a vaccine needs to build a polyclonal, durable response are weeks you have. The same patience that is a liability in advanced disease is an asset here.

TCR-T plays to the opposite conditions: established, measurable metastatic disease, where you cannot wait for an immune response to develop and need a large dose of high-avidity T cells now, and undruggable shared drivers like TP53 R175H where no small molecule exists. The price is manufacturing complexity, lymphodepletion, single-antigen escape, and HLA restriction — a given TCR works only for patients carrying the right allele. In practice the two are more likely to end up complementary than competitive: a vaccine to broaden and sustain the response, cell therapy to deliver immediate firepower, and a shared upstream engine — the prediction of which neoantigen is real — feeding both. That engine is what this site is about.

What is the difference between a neoantigen vaccine and a neoantigen-reactive TCR-T cell therapy?

Both target neoantigens — protein fragments unique to a tumor's mutations — and both rely on the same upstream step of predicting which mutated peptides are presented on the patient's HLA. They differ in delivery. A vaccine is active immunization: it shows the neoantigens to the patient's immune system so it raises its own T cells. A TCR-T is adoptive transfer: T cells are engineered to carry a receptor for the neoantigen, expanded, and infused. Vaccines are lighter to manufacture and hit many targets at once; TCR-T delivers a defined dose of high-avidity cells but is complex, single-target, and HLA-restricted.

What did ASCO 2026 show about neoantigen cell therapy?

NT-175, an engineered TCR-T therapy targeting the TP53 R175H neoantigen, reported an objective response rate of 47.6% in a first phase 1 cohort of 21 patients (Abstract 2506) — early evidence that you can build a T cell against a shared hotspot neoantigen and see tumors respond. A second program against the KRAS G12V neoantigen (Abstract 2542) was flagged as one to watch. These are early single-arm data, unlike the multi-year randomized vaccine results from intismeran and NeoVax.

What is a shared (public) neoantigen versus a personalized one?

A personalized (private) neoantigen is unique to a single patient's tumor and requires a bespoke product made per person. A shared (public) neoantigen is the same recurrent mutation found across many patients — common in genes like TP53 and KRAS — so one off-the-shelf product can treat anyone whose tumor carries it (and, for HLA-restricted cell therapies, who has the right HLA allele). NT-175 and NOUS-209 are shared-neoantigen approaches; intismeran and NeoVax are personalized.

Will neoantigen vaccines and TCR-T compete or be used together?

More likely together than head-to-head, because they suit different settings. Vaccines fit low-disease-burden settings — adjuvant therapy, minimal residual disease, prevention — where there is time to build a durable, multi-target response. TCR-T fits established metastatic disease and undruggable shared drivers, where immediate high-avidity T cells matter more than breadth. Both depend on the same neoantigen-prediction engine, so improvements there help both.