He Used AI to Build a Cancer Vaccine for His Dying Dog. It Worked.

← Back to Blog

The tumor on Rosie's leg was the size of a tennis ball.

By the time Paul Conyngham, a Sydney-based technology entrepreneur, got the full picture from his veterinarian in mid-2024, the diagnosis was devastating. Rosie — his beloved dog — had a soft tissue sarcoma that had survived both surgery and chemotherapy. The tumors were multiplying. The conventional medical options had been exhausted. The prognosis, delivered with the careful language doctors use when they have nothing left to offer, amounted to weeks, perhaps a couple of months.

Most people in that situation accept the trajectory. Conyngham, who had spent his career in computing engineering and data science, did not. What happened next — a journey from desperation to a breakthrough that leading oncologists are now calling "extraordinary" — offers one of the most remarkable examples of what happens when determined pet owners, artificial intelligence, and cutting-edge molecular biology collide.

The Diagnosis: What Rosie Was Actually Facing

Canine cancer is not a rare condition. It is, in fact, devastatingly common. According to the American Veterinary Medical Association, cancer accounts for nearly 50% of deaths in dogs over the age of 10. Approximately 6 million dogs in the United States are diagnosed with some form of cancer each year — a figure that exceeds new human cancer diagnoses annually.

Soft tissue sarcomas — the category of cancer afflicting Rosie — are particularly difficult to treat. These malignant tumors arise from connective tissue: fat, muscle, nerves, and the fibrous tissue surrounding joints. They grow locally and aggressively, are notorious for recurring after surgical removal, and are generally considered to have poor long-term outcomes even with the most aggressive conventional treatment protocols.

What made Rosie's case especially grim was the recurrence. Her cancer had already survived the standard interventions. It had demonstrated biological resilience. In oncology, that resilience is not random — it reflects the specific genetic mutations driving that particular tumor's growth. Standard chemotherapy attacks rapidly dividing cells broadly; it cannot distinguish which mutations are actually fueling the disease in any specific individual. That imprecision is precisely why recurrent cancers so often outlast the treatments thrown at them.

The AI Approach: How Conyngham Figured Out Where to Start

Paul Conyngham's background is not in biology or medicine. He holds an engineering degree, co-founded a company called Core Intelligence Technologies, and had previously served as a director of the Data Science and AI Association of Australia. He understood computation and data modeling — but the mechanics of tumor immunology were entirely outside his expertise.

He started where many people start: he asked ChatGPT. The early conversations were exploratory — queries about soft tissue sarcomas, treatment options beyond conventional chemotherapy, whether anything existed in human oncology that might apply to his dog. The AI pointed him toward immunotherapy and, critically, toward the concept of personalized cancer vaccines — a field that had been generating significant research momentum in human medicine but had almost no precedent in veterinary practice.

The core idea behind personalized cancer vaccines is elegant: every cancer is genetically unique. Tumors arise from mutations in the patient's own cells, and those mutations produce abnormal proteins — called neoantigens — that the immune system doesn't recognize as "self." A personalized vaccine can be designed to teach the immune system to identify and attack cells displaying those specific neoantigens, turning the body's own defenses into a targeted assault on the precise molecular fingerprint of that individual's cancer.

For this to work, you need to know exactly which mutations are driving the tumor. That requires sequencing — comparing the DNA of the cancer cells to the DNA of healthy cells from the same organism, identifying what changed, and then figuring out which changes matter most to the tumor's survival and growth.

The Science: AlphaFold, DNA Sequencing, and Neoantigen Mapping

Following ChatGPT's guidance toward immunotherapy research, Conyngham contacted the University of New South Wales (UNSW) Ramaciotti Centre for Genomics in Sydney. He paid $3,000 to have two samples sequenced: Rosie's healthy DNA and DNA extracted from her tumor tissue. The comparison between these two genomic profiles would reveal the exact mutations driving her cancer.

This is where another AI tool entered the picture. AlphaFold, developed by DeepMind and widely considered one of the most significant scientific breakthroughs of the decade, is a protein structure prediction system. Where previously it took years of laboratory work to determine the three-dimensional structure of a single protein, AlphaFold can predict it in hours from a genetic sequence alone. Conyngham used AlphaFold to model the abnormal proteins produced by Rosie's tumor mutations — identifying which neoantigens would be most likely to trigger a strong immune response.

The genomic data and neoantigen analysis were then brought to Pall Thordarson, director of the UNSW RNA Institute and a nanomedicine pioneer whose work has focused on mRNA delivery systems. Thordarson's team took Conyngham's computational analysis and translated it into a physical therapeutic: a custom mRNA vaccine designed to instruct Rosie's immune cells to recognize and destroy cells carrying her specific tumor proteins.

The entire development process — from DNA sequencing to vaccine synthesis — took less than two months. For context, traditional drug development pipelines for cancer therapeutics typically span a decade or more. The speed was possible precisely because AI tools had compressed the most time-consuming analytical steps from years into days.

• Rosie diagnosed with recurrent soft tissue sarcoma after surgery and chemotherapy both fail
• Conyngham begins AI-assisted research; contacts UNSW Ramaciotti Centre for Genomics
• $3,000 DNA sequencing of healthy and tumor tissue completed at UNSW
• AlphaFold used to model tumor neoantigens; key immune targets identified
• Pall Thordarson's team at UNSW RNA Institute synthesizes personalized mRNA vaccine; first injection administered
• Booster injection administered
• Results published: tennis ball-sized tumor shrunk by approximately 75%; Rosie back chasing rabbits

The Results: What Actually Happened to Rosie

Rosie received her first injection of the personalized mRNA vaccine in December 2024. A booster followed in February 2025. The response, documented by the research team at UNSW, exceeded what even optimistic projections had anticipated.

The primary tumor on her leg — the tennis ball-sized mass that had survived both surgery and chemotherapy — shrank by approximately 75%. Multiple other tumor sites also demonstrated measurable regression. By March 2026, when the story broke internationally in outlets including Fortune, Newsweek, and the Sydney Morning Herald, Rosie was described by Conyngham as "back chasing rabbits" — mobile, energetic, and responding to life with the enthusiasm of a dog that had not recently been given weeks to live.

Dr. Pall Thordarson, speaking to multiple outlets, called the result remarkable not just as an individual animal's recovery but as a proof-of-concept for the entire approach. This was the first time a fully personalized cancer vaccine had ever been designed for and administered to a dog. The experimental framework Conyngham assembled — using consumer-accessible AI tools to navigate specialized scientific literature and connect with expert researchers — represents a genuinely new model for how determined patients and pet owners might engage with cutting-edge medicine.

What This Means for Every Dog Owner

The medical community's response to Rosie's story has been cautiously optimistic. Oncologists have been careful to note that a single successful case — however dramatic — does not constitute clinical evidence. One animal's response, however remarkable, cannot confirm efficacy, safety profiles, or reproducibility across a population. Randomized controlled trials, with all their complexity and expense, remain the gold standard for establishing that any medical intervention actually works consistently.

What researchers are excited about, however, is the pathway. The tools Conyngham used — ChatGPT for literature navigation, AlphaFold for protein structure prediction, next-generation genomic sequencing, mRNA synthesis — are all available today. The technical infrastructure for personalized cancer vaccine development in veterinary medicine exists. What has been missing, until now, is the assembled proof that it can work in a living patient.

For dog owners facing a cancer diagnosis in their pet, the story raises a practical question: could this approach become accessible? The honest answer is that personalized mRNA cancer vaccines for companion animals are not yet a standard treatment option, and may not be for several years. The $3,000 sequencing cost alone places it outside reach for many families, and the need for a specialized research institution to synthesize the vaccine means this is not something a veterinary clinic could replicate tomorrow.

But the timeline is compressing fast. Human clinical trials for personalized mRNA cancer vaccines — including a high-profile partnership between Moderna and Merck for melanoma — have already demonstrated meaningful results. The same wave of investment and infrastructure development that is accelerating personalized cancer medicine for humans is beginning to ripple into veterinary oncology. Rosie may be the first dog to receive a personalized AI-designed mRNA cancer vaccine. She is almost certainly not the last.

What to Do If Your Dog Is Diagnosed With Cancer Today

For dog owners reading this after a cancer diagnosis, Rosie's story offers both inspiration and a practical framework. You do not need to be a technology entrepreneur or have connections at a research university to advocate effectively for your animal. But you do need to be an active, informed participant in the process.

Seek a board-certified veterinary oncologist, not just a general practice vet, for any cancer diagnosis. Ask specifically about clinical trials — the Veterinary Cancer Society and the National Canine Cancer Foundation both maintain searchable directories of ongoing research trials, many of which offer access to cutting-edge treatments at reduced or no cost in exchange for data collection. Ask your oncologist directly: "Is there a clinical trial my dog might qualify for?"

Document everything. Get copies of all pathology reports, imaging results, and biopsy findings. This information is the foundation for any second opinion, any research inquiry, or any conversation with a specialist. Knowing the specific histological subtype of a tumor, the margins of a surgical resection, and the grade of the malignancy are all essential data points that can change the direction of treatment.

And — if you have the inclination and the technical curiosity — consider the path Conyngham walked. AI tools like ChatGPT are genuinely useful for navigating medical literature, identifying research directions, and formulating the right questions to ask specialists. They are not a replacement for medical expertise, but as Rosie's story demonstrates, they can be a remarkable bridge between a determined owner and the experts who hold the pieces of a solution.

The Bigger Picture: AI and the Future of Veterinary Medicine

Rosie's story sits at the intersection of several forces reshaping medicine simultaneously: the democratization of AI tools, the plummeting cost of genomic sequencing, the maturing infrastructure of mRNA technology, and the growing recognition that companion animals deserve access to the same quality of precision medicine that humans increasingly expect.

The bond between humans and their dogs is ancient and well-documented in the archaeological record — stretching back at least 15,000 years. What has changed is the sophistication of what we can now offer these animals when their health fails. The gap between what is theoretically possible and what a devoted owner with access to the right tools can actually achieve for their pet has narrowed dramatically in the span of just a few years.

Paul Conyngham didn't set out to advance veterinary oncology. He set out to save his dog. In doing so, he may have done both. And somewhere in Sydney, a medium-sized dog with a complicated medical history is apparently chasing rabbits again — which, in the end, is what all of this was for.