Personalized mRNA Cancer Vaccines: How Far Along Are They, Really?

Why this question is being asked now

In March 2026 a story travelled further than most cancer research does. An Australian tech entrepreneur, Paul Conyngham, had used AI tools and a university laboratory to help design a one-off mRNA vaccine against his dog's cancer, and several of the dog's tumors shrank. Sam Altman shared it. Short-video versions of it are still circulating.

Five months later, in August 2026, two things happened in the same week. Conyngham launched a company to build these vaccines for dogs, and its launch materials said his dog Rosie's cancer had returned after a later surgery. Separately, Merck and Moderna announced that a large randomized trial of a personalized mRNA vaccine in melanoma had met its main goals.

One of those is an anecdote. One is evidence. Both are about the same underlying idea, and telling them apart is the whole point of this article.

The idea in plain language

Cancer starts when the instructions inside a cell get changed. DNA is a very long string of chemical letters. When enough of those letters are wrong, in the wrong places, a cell can start growing and dividing without stopping.

Those errors have a side effect. Some of them cause the cell to build proteins that are slightly wrong - proteins that exist nowhere else in the body. NCI calls these neoantigens, meaning new antigens: substances the immune system has never seen and did not learn to ignore.

That is the opening. The immune system spends its life distinguishing self from not-self, and it usually files cancer cells under self, because they are the body's own cells. A neoantigen is a piece of the cancer that does not look like self. If you could point the immune system at it, the system might do the rest.

A cancer treatment vaccine is that pointing. NCI is careful about the word, because two very different things share it:

• A prevention vaccine works against something that causes cancer - the HPV vaccine and the hepatitis B vaccine both work this way. They are given before there is cancer.
• A treatment vaccine is given to someone who already has cancer, and it targets the cancer cells themselves.

Everything in this article is the second kind. Confusing the two is the single most common error in coverage of this field.

Where mRNA comes in

If DNA is the master instruction book, messenger RNA is a photocopy of one page, sent to the part of the cell that builds proteins. It carries one command: make this protein. The cell makes it, and then the mRNA is broken down. It does not enter or change your DNA.

That property - a short-lived instruction that gets a specific protein built on demand - is what makes mRNA useful here. Instead of manufacturing a neoantigen protein in a factory and injecting it, you inject the instruction, wrapped in a tiny fat bubble called a lipid nanoparticle, and let the person's own cells build it. Their immune cells encounter the fragment, learn it, and go looking for anything else displaying it.

The COVID mRNA vaccines used the same delivery method with a different message. That is the extent of the relationship. A personalized cancer vaccine is not a COVID vaccine, is not given to healthy people, and is not preventive.

What "personalized" actually costs you

Here is the sequence that has to happen for one patient:

1. Take tumor tissue and healthy tissue. You need enough tumor. Some patients do not have enough, or cannot safely have another biopsy.
2. Sequence both, and subtract. Reading the tumor's DNA and the patient's normal DNA and comparing them produces a list of mutations unique to the cancer. This is the same tumor testing described in biomarker testing and precision medicine.
3. Predict which mutations matter. Most mutations are noise. Of the rest, only some produce a protein fragment that the patient's particular immune system can actually display and recognize. This prediction step is computational, and it is where AI tools have made the largest difference.
4. Design and manufacture the mRNA. Usually encoding several chosen neoantigens at once, then formulate it in lipid nanoparticles.
5. Give it, usually alongside something else. In most trials the vaccine is added to a checkpoint inhibitor - an immunotherapy that releases a brake the tumor uses to shut down immune attack. The vaccine points; the checkpoint inhibitor permits.

Every dose is built for one person. Nobody else's would work. That is the promise and the bottleneck at the same time: no economies of scale, a manufacturing run per patient, and weeks between biopsy and first dose - weeks the patient's cancer does not pause for.

What the human evidence shows

Small trials, strong immune signals. NCI reported in April 2025 on two early trials of personalized neoantigen vaccines. At Dana-Farber, nine patients with kidney cancer were treated after surgery; NCI reported no recurrences, with four patients past three years cancer-free, and immune cells that still recognized their tumor's neoantigens years later. At Memorial Sloan Kettering, sixteen patients with pancreatic cancer were treated; eight mounted a strong immune response, and six of those remained cancer-free at around three years, while non-responders relapsed at a median of about thirteen months. Side effects in both were mild.

Those are real results, and NCI reported them with the caveats attached: the trials were tiny, larger ones were already underway to confirm them, and in kidney cancer long remission after surgery is common regardless of treatment, so the vaccine cannot simply be credited with it.

One large randomized trial, results announced but not published. In August 2026 Merck and Moderna said their Phase 3 trial in melanoma, INTerpath-001, showed that adding a personalized mRNA vaccine to Keytruda kept the cancer from returning longer than Keytruda alone, in patients whose melanoma had been surgically removed. That is the first result of this kind at that scale. The announcement did not include the actual numbers, the trial is ongoing, and the vaccine is not approved or available outside a trial. Our page on the INTerpath-001 result goes through what was and was not said.

What has actually finished the road. NCI lists two approved treatment vaccines: sipuleucel-T, for advanced prostate cancer that no longer responds to hormone therapy, and T-VEC, for melanoma that comes back after surgery. Neither is a personalized neoantigen mRNA vaccine. Everything in that category is still investigational, and the only legitimate route to one is a clinical trial.

Side effects are not zero. NCI names flu-like symptoms as common with treatment vaccines - fever, chills, weakness, dizziness, nausea. Severe reactions are possible, and vary by product.

So what did the dog case add?

Scientifically, close to nothing, and the people involved said so. It was one animal, treated outside any trial, alongside a checkpoint inhibitor, with a cancer type - canine mast cell tumor - known to behave unpredictably. Tumor grade and other clinical details were never published. There was no comparison animal. Some tumors never responded. The cancer later came back.

What it did demonstrate is a change in access. Justin Stebbing, an oncologist writing in The Conversation, framed it as the ingredients of high-end personalized medicine becoming reachable: a motivated person can now pay for tumor sequencing, use AI to help interpret the results, and partner with an academic lab to turn that interpretation into an mRNA construct. The AI did not cure anything. It acted as an always-available guide, and trained scientists at UNSW - including its RNA Institute - reviewed the design and did the laboratory work.

Stebbing also named the problem that follows: the field now needs ways to test approaches like this properly, to protect patients and animals from false hope and unsafe experiments, and to decide who gets access if they work. Cheap capability without a testing framework is how people get hurt.

The honest open questions

• Does personalization earn its cost? It is not yet established that a bespoke vaccine beats simpler, non-personalized approaches for the same cancer. That comparison has largely not been run.
• Which neoantigen predictions are right? Choosing targets is still partly guesswork, and a vaccine aimed at fragments the immune system cannot display does nothing.
• How much is the vaccine and how much is the checkpoint inhibitor? Most trials give both. Attribution requires designs that separate them.
• Does it work in advanced disease? The strongest results so far are in the setting after surgery, where the goal is preventing return rather than shrinking established tumors.
• Who could ever afford it? A per-patient manufacturing model at scale is an unsolved economic problem, not just a scientific one.

Reading the next headline

The pattern repeats often enough to be worth naming. A striking case appears, the AI angle carries it far beyond the medical press, and the caveats fall off in transmission. A few questions filter most of it:

• Is this a single case or a trial with a comparison group?
• How many people, and for how long were they followed?
• Were the numbers published, or only announced?
• What else were the patients receiving at the same time?
• Is the treatment available outside a trial? For everything in this field today, the answer is no.

None of that is a reason for cynicism. The kidney and pancreatic results are genuinely encouraging, and a positive Phase 3 in melanoma is a real milestone. It is a reason to hold the hope and the evidence at the same time, which is harder than holding either one alone.

Questions worth asking a healthcare team

• Has my tumor been tested in a way that would make me eligible for a vaccine trial?
• Is there an open trial for my cancer type and stage, and where is it running?
• What would I be giving up, if anything, by joining one?
• What is the standard treatment I would otherwise receive, and what does it offer?
• How would we know whether it was working?

How this article was prepared

An AI-assisted editorial system helped prepare this page. No named medical reviewer has reviewed it unless one is listed.

The National Cancer Information Foundation publishes Cancer Explained. This page is for learning. It is not medical advice and does not suggest a test or treatment. It is not veterinary advice.

See an error, old source, or unclear wording? Tell us.