The missing chapter: personalized neoantigen vaccines, and why hemangiosarcoma is a hard test.

What a personalized neoantigen vaccine is

Every cancer starts as a normal cell that picked up mutations. Some of those mutations change the shape of the proteins the cell makes, and a changed protein fragment can look foreign to the immune system in a way the healthy version never did. Those altered fragments have a name: neoantigens, literally "new antigens." They exist only in the tumor, not in any healthy cell, which makes them the cleanest possible target. The immune system, if it can be pointed at them, can attack the cancer while leaving normal tissue alone.

A personalized neoantigen vaccine is built to do exactly that pointing. The recipe, in plain terms, is four steps:

1. Sequence the tumor. Read the DNA of the dog's own tumor and compare it to the dog's healthy DNA. The differences are the tumor's private mutations.
2. Predict the good targets. Software works out which of those mutations produce a fragment the dog's own immune system can actually display and recognize. Not every mutation makes a usable target, so this step ranks the candidates.
3. Write the blueprint. The top candidates are encoded into a strip of mRNA, the same delivery technology behind the recent human vaccines. The mRNA is a set of instructions telling the dog's cells to make those exact fragments briefly, so the immune system can learn them.
4. Teach the immune system. The vaccine is given, the dog's cells display the fragments, and the immune system builds a response aimed at cells carrying those specific mutations, which is to say, the tumor.

The whole point is in the first word: personalized. There is no single RosieVaccine shot sitting on a shelf. Each design is made from one dog's tumor, for that one dog.

How it differs from the vaccines in Part 3

In Part 3 we walked the scoreboard of dog cancer vaccines. It is worth seeing where this approach sits relative to those, because the differences are the whole argument.

The off-the-shelf approaches have a real advantage: they are ready immediately, no sequencing required. Their limit is that they aim at targets chosen in advance, so they can only help dogs whose tumors happen to carry those targets. The whole-tumor approach is individual to the dog but hands the immune system the entire tumor at once, with no control over which parts it learns from. The personalized neoantigen approach tries to get the best of both: individual to the dog, like the whole-tumor vaccines, but selective, aiming only at the specific mutated fragments most likely to provoke a response. It is the most tailored of the four, and, as we will see, also the most demanding to build.

The proof it can work, so far in people

Does this actually change outcomes, or is it just an elegant idea? The strongest evidence to date comes from human medicine. In a randomized trial in people whose high-risk melanoma had been surgically removed, adding a personalized neoantigen vaccine (mRNA-4157, also called V940, from Moderna and Merck) to the immunotherapy drug pembrolizumab lowered the risk of the cancer returning, or of death, by about 44 percent compared with pembrolizumab alone (Weber et al., The Lancet, 2024).

That is a genuinely important result. It is the clearest sign yet that reading one patient's tumor and building a bespoke vaccine from it can move survival, not just stir up an immune response in a lab test. But two honest caveats belong right next to it. First, this is human data, in melanoma, and a result in people does not automatically transfer to dogs or to a different cancer. Second, melanoma was chosen for a reason: it is one of the most heavily mutated cancers there is, which gives a neoantigen vaccine a long list of targets to work with. Hemangiosarcoma is not like that, and that is where the honesty of this post has to sharpen.

Why hemangiosarcoma is a hard test

If a personalized neoantigen vaccine feeds on mutations, then the ideal cancer for it is one with many mutations. Hemangiosarcoma is, by the current evidence, a modestly mutated cancer with few recurring drivers, and that makes it one of the harder cancers to build a neoantigen vaccine for, not one of the easier ones. We would rather tell you that plainly than let the elegance of the idea oversell it.

Here is the single clearest piece of data. When researchers sequenced 27 hemangiosarcoma tumors in detail, they could identify a likely cancer-driving mutation in only about half of them (Wong et al., PLoS One, 2022). Roughly half the tumors had no clear driver at all.

Three things flow from that picture, and each is a real headwind:

• Fewer mutations means fewer targets. A modestly mutated tumor gives the design software a shorter list of usable neoantigens to choose from. This is the opposite of the melanoma setting where the approach was first proven.
• Tumors are not uniform inside. The same study found that different regions of a single hemangiosarcoma tumor carried different mutations, an evolving, branching family tree rather than one clone. A vaccine built from one piece can miss mutations living in another piece.
• The starting material is often difficult. Many tumor samples are preserved in a way (formalin fixation, the standard for pathology) that chemically damages the DNA and makes confident mutation-calling harder. We wrote a whole post on how sample preservation changes what is possible, precisely because it matters this much.

Where that leaves us, honestly

So why work on it at all, if hemangiosarcoma is such a hard case? Two reasons, and we hold both at once.

The first is that the personalized approach does not need a recurring, shared driver mutation, and that is exactly the gap the other approaches leave open. Targeted drugs need a known druggable driver, which, as the chart shows, only about half of these tumors have. Off-the-shelf vaccines need their pre-chosen target to be present. A vaccine read from the individual tumor can, in principle, use that tumor's own private mutations, whatever they happen to be. For a cancer with few shared drivers, an individualized strategy is a rational answer rather than a fashionable one.

The second reason is simply that the honest headwinds above are engineering problems, not dead ends. Careful paired sequencing of tumor against healthy tissue, sampling that accounts for a tumor's internal variety, and artifact-aware handling of difficult preserved samples are all things a well-designed pipeline can be built around. That is the work.

To be completely clear about where this stands: a personalized neoantigen vaccine for dogs is an approach under active development, not a proven or approved therapy, and nothing here is a claim that it extends survival. In the United States, therapeutic cancer vaccines for animals are overseen by the USDA Center for Veterinary Biologics, not the FDA, and a real product has to earn its evidence the same careful way every treatment in this series did. What we can say plainly is that this is the one chapter the rest of the field has left mostly unwritten for hemangiosarcoma, and that filling it in responsibly is worth doing.

The whole series in one view

Five posts, one throughline: hemangiosarcoma is a hard cancer, the tools are real but none is a cure alone, and the strength of the evidence is what separates hope from hype.

If you have read all five, you now know more about this disease and its treatment landscape than most dog owners ever get the chance to. Take that into the room with your oncologist and ask better questions.

A note on sources

This series was prompted by a public educational webinar from the Canine Cancer Alliance, a nonprofit that funds canine cancer research. Peer-reviewed claims are cited to their original studies by DOI or PubMed. Where the evidence is human-only, we say so plainly, because with this cancer the strength of the evidence is the whole story.