AI-Designed mRNA Vaccine Shrinks Dog's Cancer Tumor

"We took her tumour, sequenced the DNA, we converted it from tissue to data, and we used that to find the problem in her DNA and then develop a cure based off that. ChatGPT assisted throughout that entire process."
Paul Conyngham

A Sydney tech entrepreneur with no biomedical degree used ChatGPT to plan his dying dog's DNA sequencing, AlphaFold to model the resulting protein structures, and an Australian university lab to produce a custom mRNA vaccine - all within two months. The tumor on his rescue dog Rosie's leg shrank by 75%. Scientists involved say it is the first personalized cancer vaccine ever designed for a dog.

TL;DR

What they said: AI-designed mRNA vaccine produced the first personalized cancer treatment for a dog, shrinking a mast cell tumor by 75%
What we found: The science checks out - real institutions (UNSW RNA Institute, University of Queensland), real ethical approvals, real tumor response - but it is a single anecdotal case, not clinical evidence
Paul Conyngham used ChatGPT for treatment planning and target identification, AlphaFold for protein structure prediction, and machine learning for neoantigen selection
Vaccine produced at UNSW, administered at University of Queensland in December 2025; visible results within one month
Scientists are cautiously excited but stress this needs controlled trials before drawing broader conclusions

The Claim

Paul Conyngham, a 17-year veteran of machine learning and data science who runs Sydney consultancy Core Intelligence, adopted Rosie - a Staffordshire Bull Terrier-Shar Pei cross - from a shelter in 2019. In 2024, large tumors appeared on one of her back legs. The diagnosis: mast cell cancer, the most common skin cancer in dogs and normally incurable through conventional treatment.

Conyngham tried everything conventional veterinary medicine offered. Multiple surgeries. Chemotherapy. Immunotherapy. The chemo slowed the tumor spread but failed to shrink the growths. Vets estimated Rosie had between one and six months to live.

That is when Conyngham turned to the tools he knew. He started with ChatGPT - not as a medical device, but as a research assistant. The model suggested immunotherapy as a direction and pointed him toward genomic sequencing. It helped him plan the pipeline: sequence Rosie's tumor DNA, compare it to her healthy cells, identify the mutations driving the cancer, and design a vaccine targeting those specific neoantigens.

The Evidence

The AI Pipeline

The technical workflow combined three AI systems:

ChatGPT - Research planning, treatment strategy, and iterating on vaccine design. Conyngham used it all through the process to brainstorm approaches and analyze genetic data
AlphaFold - Google DeepMind's protein structure prediction tool, used to model the 3D structures of proteins encoded by Rosie's tumor mutations
Custom machine learning - Conyngham's own algorithms for neoantigen selection, identifying which mutated proteins would most likely trigger an immune response

The output: a half-page formula describing the mRNA sequence for a vaccine targeting Rosie's specific cancer mutations.

The Institutions Are Real

This was not a garage operation. Conyngham brought his data to UNSW's Ramaciotti Centre for Genomics, where researchers initially hesitated but agreed to collaborate after reviewing his analysis. Pall Thordarson, director of the UNSW RNA Institute, designed and produced the bespoke mRNA vaccine.

"Once we had the sequence that Paul designed, it was less than two months until we handed it over to Paul, to the vet," Thordarson said. He called it "the first time a personalized cancer vaccine has been designed for a dog."

Mari Maeda of the Canine Cancer Alliance connected Conyngham with Rachel Allavena at the University of Queensland, who had the ethical approvals required to administer experimental veterinary treatments. Rosie received her first injection in December 2025, followed by a booster shot.

A dog receiving veterinary treatment Mast cell cancer is the most common skin cancer in dogs and is typically incurable through conventional treatment. Conyngham's AI-designed vaccine represents the first personalized approach. Source: pexels.com

The Results

Within around one month, Rosie's tumor shrank by 75%. By January 2026, Conyngham watched her jump over a fence to chase a rabbit - behavior that had disappeared during her illness.

Martin Smith, associate professor of computational biology at UNSW, did not hide his reaction: "It was like holy crap, it worked!"

David Thomas, inaugural director of UNSW's Centre for Molecular Oncology, called it "a very impressive thing," highlighting the citizen science angle - a non-academic leveraging publicly available AI tools to do work that would normally require a specialized research team.

Kate Michie, UNSW structural biologist, said it's "encouraging that a non-scientist could execute such a pipeline."

What They Left Out

This Is One Dog

The most important caveat: this is a single anecdotal outcome, not clinical evidence. One tumor in one dog responded to one vaccine. The result is encouraging but proves nothing about reproducibility, safety across a population, or long-term durability. Controlled trials with proper sample sizes, control groups, and blinded evaluation do not exist yet.

Conyngham himself is already working on a second vaccine targeting another tumor that did not respond to the initial treatment. That a second tumor on the same dog didn't shrink is itself a data point about the limitations of the approach.

The AI Did Not Design the Vaccine Alone

The headline framing - "AI creates cancer vaccine" - overstates what happened. ChatGPT served as a research assistant and planning tool. AlphaFold predicted protein structures. But the critical decisions - which neoantigens to target, how to design the mRNA sequence, and how to confirm the design - required human judgment from Conyngham and deep domain expertise from Thordarson's team at UNSW.

The AI accelerated a pipeline that would otherwise require months of manual literature review and computational modeling. It didn't replace the immunologists, the RNA engineers, or the veterinary oncologists who made the vaccine real.

Regulatory Reality

Rosie's treatment was possible because veterinary experimental treatments face lighter regulatory scrutiny than human medicine. There's no veterinary equivalent of FDA Phase I-III clinical trials for a one-off compassionate use case. Scaling this to a standardized treatment - for dogs or humans - would require years of regulatory work that this single case doesn't shortcut.

Component	What Happened	What's Missing
AI-assisted design	ChatGPT + AlphaFold + custom ML	Peer-reviewed validation of target selection
Vaccine production	UNSW RNA Institute (2 months)	Scalable manufacturing process
Administration	UQ with ethical approval	Controlled clinical trial
Tumor response	75% shrinkage in one tumor	Response in other tumors, long-term followup
Reproducibility	Single case (n=1)	Multi-animal trial with controls

The story of Rosie and Paul Conyngham is real, verified, and truly moving. A dog owner with the right technical background used publicly available AI tools to do something that has never been done before - design a personalized cancer vaccine for his pet, produced and administered by legitimate research institutions, with a measurable tumor response. The scientists involved are cautiously excited. But the distance between "one dog's tumor shrank" and "AI can cure cancer" is measured in years of clinical trials, regulatory approvals, and reproducibility studies. What this case does prove is that the AI-assisted pipeline works fast enough and cheaply enough that a motivated individual can execute it. That speed - from diagnosis to injection in months, not years - is the part worth watching.

Sources: