Personalized vs. Shared-Antigen Cancer Vaccines for Dogs: What's the Difference?

Three strategies, one goal

Every cancer vaccine, human or animal, wants the same thing: to train the immune system to recognise cancer cells and destroy them. Chemotherapy poisons dividing cells. Radiation burns them. A vaccine instead teaches the body's own patrol system what to look for. That branch of medicine is called immunotherapy.

Where the strategies part company is in what they ask the immune system to look for. There are three broad answers, and the differences between them are not cosmetic. They lead to different costs, different timelines, different safety profiles, and different answers to the only question that matters to you: is this right for my dog?

1. A shared antigen. Pick one marker that many tumors carry. Make one product. Give it to everybody with that cancer type.
2. The whole tumor. Do not pick anything. Use the dog's own tumor tissue as the source material and let the immune system sort out what to attack.
3. Personalized neoantigens. Sequence the individual tumor, find the mutations that exist only in it, and build a vaccine around those.

An antigen, throughout this article, just means "something the immune system can recognise and target."

Strategy 1: the shared antigen

A shared-antigen vaccine targets a protein that many tumors of a given type express, or that cancer cells over-produce compared with healthy cells. The target is chosen in advance, built into a product manufactured at scale, and given to any patient whose tumor type is likely to carry it. It is off-the-shelf: designed once, deployed broadly.

The economics are excellent. One product can go through a single regulatory review, sit in a freezer at any oncology clinic, and be given on the day of diagnosis with no waiting for anything to be custom-built.

The biology is where it gets hard. A protein the tumor shares with normal tissue is, by definition, a protein the immune system has already met and decided to ignore. That deliberate ignoring is called self-tolerance, and it is not a bug. It is the mechanism that stops your dog's immune system from attacking your dog. Asking the immune system to break tolerance against a self-protein means asking it to do the one thing it was carefully trained not to do. That is difficult, the resulting response tends to be modest, and it carries an inherent ceiling: any attack on a shared target has to somehow spare the healthy cells that also carry it.

What ONCEPT taught the field

The landmark example in veterinary medicine is ONCEPT, a canine melanoma vaccine developed from the work of Philip Bergman and colleagues. When it was licensed by the USDA, it became the first therapeutic cancer vaccine licensed for use in any species, human or animal. That is a genuine milestone, and veterinary oncology got there before human oncology did.

ONCEPT delivers the gene for human tyrosinase, an enzyme that pigment cells use to make melanin. Because the human version differs slightly from the dog's own, the dog's immune system treats it as mildly foreign, and the response can spill over onto the dog's melanoma cells. It is a clever way to get around self-tolerance: do not present the dog's protein, present a close relative of it.

It is also a good illustration of the ceiling. Tyrosinase is a shared self-antigen, present in every healthy pigment cell the dog has. The published clinical evidence for ONCEPT is mixed enough that veterinary oncologists still actively debate how much survival benefit it delivers in practice. We cover that debate in our guide to melanoma in dogs.

The newest shared-antigen bet: VACCS

The most ambitious shared-antigen effort in dogs is the Vaccination Against Canine Cancer Study (VACCS), run by Colorado State University with the company Calviri. It is worth understanding because it is a genuinely different bet, and because it is easy to misread.

VACCS does not treat cancer. It tries to prevent it. Healthy, cancer-free dogs received a single off-the-shelf vaccine built from a fixed panel of 31 frameshift neoantigens. A frameshift is a particular kind of DNA error that knocks the reading of a gene out of alignment, producing a stretch of protein that is wildly wrong rather than subtly wrong. The insight behind VACCS is that some of these frameshift products recur across many tumors and even across species, which would make them shared targets that are still genuinely abnormal.

More than 800 dogs were enrolled, making it the largest interventional cancer trial ever conducted in companion dogs. The primary question is whether vaccinated dogs go on to develop fewer cancers of any type. (Burton et al., Veterinary Immunology and Immunopathology, 2023)

Here is the part to be careful about, and we would rather say it plainly than let you infer it: the trial finished enrolling and following dogs, but the results have not been published. Nobody yet knows whether shared frameshift prevention works in dogs. Anyone who tells you it does, or that it does not, is ahead of the evidence. We are watching for that readout like everyone else.

Strategy 2: use the whole tumor

The second strategy sidesteps target selection entirely. Instead of choosing which molecule to aim at, you take the dog's own surgically removed tumor, process it, and give it back as the vaccine. Because the material comes from that individual animal, this is called an autologous approach.

In veterinary oncology the best-known example is ELIAS Animal Health, which pairs a vaccine made from a dog's own tumor tissue with a second step in which the dog's T-cells (the immune system's patrol and killer cells) are grown outside the body and infused back in. Their osteosarcoma product received USDA Center for Veterinary Biologics approval in 2025.

The appeal is obvious. Everything unique about that tumor is in the vaccine, including targets nobody would have thought to predict. No sequencing, no algorithms, no guessing.

The cost is control. Because you never chose the targets, you cannot know what the immune system settled on, and you cannot exclude anything. The tumor tissue also contains the dog's ordinary healthy proteins, which are along for the ride. There is no step at which a candidate target can be inspected and rejected. And because it depends on physical tissue rather than data, it cannot be designed remotely, shipped as a file, or reproduced later.

Strategy 3: the personalized neoantigen vaccine

The third strategy is the one RosieVaccine works on, and it tries to keep what is good about the other two: the target selection of the shared-antigen approach, and the patient-specificity of the autologous approach.

Instead of picking a target that many tumors share, or refusing to pick at all, you sequence the individual tumor, identify the mutations found only in it, and build the vaccine from those. It cannot be stockpiled. A vaccine designed from one dog's tumor is useless to any other dog. That is not an inconvenience of the method. It is the method.

So what exactly is a neoantigen?

Cancer is a disease of mutations, which are simply typos in a cell's DNA. Some typos are silent and change nothing. Some swap a single building block in a protein, so the cancer cell builds a protein that is very slightly wrong.

Cells constantly chop up the proteins they are making and hold the fragments up on their surface, like showing samples of everything happening inside. Patrolling T-cells read those fragments. When a cancer cell displays a fragment of one of its misspelled proteins, it is displaying something that exists nowhere in the healthy animal.

That fragment is a neoantigen. "Neo" for new, "antigen" for something the immune system can target. It is genuinely foreign, in the same sense a virus is foreign, even though the dog's own cell built it.

This is why the distinction from a shared antigen is not academic. There is no self-tolerance to overcome, because the immune system has never seen this thing before and was never taught to leave it alone. And a target that appears nowhere in healthy tissue is a target that cannot, in principle, direct the immune system against healthy tissue.

That specificity is simultaneously the safety story and the efficacy story. They are the same property seen from two sides. It is also why the comparison against the dog's healthy DNA is not bookkeeping: it is the safety step.

How a personalized vaccine gets designed

It starts with two samples, not one. A piece of the tumor, and a sample of healthy tissue, usually a simple blood draw.

Every dog is born with thousands of harmless quirks in their DNA, present in every cell they have. Those inherited variants are called germline. Comparing the tumor against the dog's own healthy DNA strips all of those away, leaving only the changes the tumor acquired, which are called somatic mutations. Without the matched healthy sample you cannot tell the two apart, and you would risk designing a vaccine against the dog's own inherited proteins. This is a safety requirement, not a preference. (If your dog is heading for surgery, this is also why how the tumor sample is preserved matters so much.)

Next, not every somatic mutation produces a neoantigen the immune system can actually see. To be read by a T-cell, a fragment must first be held up in a molecular display case called MHC. In dogs that display system is DLA (Dog Leukocyte Antigen); in people it is called HLA. Different dogs inherit differently shaped display cases, so a fragment that slots neatly into one dog's may not fit another's at all. Software predicts which candidates will fit this dog's, producing a ranked shortlist.

Those chosen fragments are then written into a single mRNA sequence. mRNA is the body's short-term instruction note: a cell reads it, builds what it says, and discards it. It never enters the dog's DNA and it does not persist. When the vaccine is given, the dog's own cells read the note, build those exact fragments, display them, and the immune system learns what to hunt.

Two things are worth stating clearly. The vaccine contains no cancer and cannot cause cancer. And prediction is prediction: an algorithm's confidence that a fragment will be displayed is a probability, not a certainty. The leading human programmes confirm afterwards, by measuring the patient's actual T-cell response, whether the predictions were right. That confirmation step is a real and honest gap between prediction-based design and the human research frontier.

The honest trade-off

No strategy dominates. Each buys something and pays for it somewhere else.

Put simply: personalized gives you maximal specificity and the cleanest safety argument, and charges you in sequencing cost and weeks of turnaround. Shared is cheap and scalable and less specific. Whole-tumor needs no prediction and gives you no control over what gets targeted.

Where the evidence actually stands

This is the section where it would be easy to oversell, so we are going to be conservative on purpose.

In humans, the mechanism is proven and the efficacy signal is real but early

The strongest evidence is mRNA-4157 (also called V940) from Moderna and Merck. In a randomised trial in people whose high-risk melanoma had been surgically removed, adding a personalized neoantigen vaccine to the immunotherapy drug pembrolizumab lowered the risk of the cancer returning, or of death, by about 44 percent compared with pembrolizumab alone. Longer follow-up at around three years put that figure near 49 percent. (Weber et al., The Lancet, 2024)

The caveats belong right beside the number. That was a phase 2b trial with 157 patients, which is small, and the headline result sat just at the edge of conventional statistical significance. A much larger phase 3 trial is running now to settle it. This is a strong, consistent signal. It is not a settled fact.

A second programme, autogene cevumeran from BioNTech and Genentech, tested the same architecture after surgery in pancreatic cancer. In a phase 1 trial of 16 patients, it induced new neoantigen-specific T-cells in half of them, and those T-cells were still detectable years later. (Rojas et al., Nature, 2023)

In dogs, nobody has a licensed personalized neoantigen vaccine

That is the plain truth, and it includes us. There is no canine equivalent of the melanoma trial above.

The closest and most encouraging canine work comes from Elias Sayour, Duane Mitchell and colleagues at the University of Florida, published in Cell in 2024. They built personalized mRNA vaccines for client-owned dogs with terminal brain tumors and reported improved survival, then carried the same technology into a first-in-human glioblastoma trial. (Mendez-Gomez et al., Cell, 2024)

Two honest qualifications. Their vaccines were made from the tumor's own RNA rather than from computationally predicted neoantigens, so it is a cousin of our approach rather than the same thing. And those dogs had gliomas, a brain cancer, not the solid tumors most owners are reading this about. What it does establish, and this matters, is that a personalized mRNA cancer vaccine can be made for an individual dog and can do something.

So the fair summary for 2026: the mechanism is established, the human efficacy signal is genuine but still maturing, and in dogs this remains an emerging, experimental approach whose expected benefit is extrapolated from human data rather than demonstrated in a canine trial. That extrapolation is a hypothesis. It is not evidence.

What this means for your dog

The question is almost never "which vaccine strategy?" in isolation. It is what combination of treatments gives this particular dog the best chance. Local control comes first, meaning surgery, radiation, or both. Immunotherapy is generally given afterwards, in what oncologists call the adjuvant setting, to reduce the chance the cancer returns. Every human trial described above was run that way.

The strategies are not rivals in any simple sense. A dog with oral melanoma might reasonably receive ONCEPT. A dog with osteosarcoma might be a candidate for an autologous product. A dog with a solid tumor carrying enough mutations might be a candidate for a personalized design. These are conversations for a veterinary oncologist who has examined your dog, has the pathology in front of them, and knows the timeline you are working against.

One practical constraint is worth knowing in advance. A personalized vaccine depends on getting enough good-quality DNA out of the tumor. Very small samples, heavily necrotic samples, and tissue that has been sitting in formalin can all make that harder. It is usually workable, but it is a conversation to have early rather than late.

Where RosieVaccine fits, and what we are not

We are a computational vaccine design service. We take a tumor sample and a matched blood sample, identify the mutations unique to that dog's cancer, predict which of them the dog's immune system can display, and design an mRNA blueprint targeting them. That blueprint goes to a manufacturing partner. A licensed veterinarian orders the manufacturing, administers the vaccine, and remains in charge of your dog's care from beginning to end. We do not manufacture vaccines, we do not treat patients, and we do not practise veterinary medicine.

We want to be equally clear about the limits. This is an experimental approach. It is not an FDA-approved drug. In the United States, veterinary biologics like this fall under the USDA Center for Veterinary Biologics, not the FDA. It is designed to train the immune system. We cannot promise a cure, a remission, or any particular outcome for your dog, and you should be sceptical of anyone who does.

What we can offer is a design built on the same architecture that produced the strongest results in human oncology, an honest account of what is known and what is not, and a licensed veterinarian standing between our software and your dog.