Paul Conyngham is not a biologist or veterinarian, but an Australian engineer with nearly twenty years of experience in data science and artificial intelligence. In 2024, his dog Rosie was diagnosed with mast cell cancer, the most common type in canine skin and difficult to treat with conventional therapies. After exhausting standard options, Conyngham turned to AI: he opened ChatGPT and began researching. ChatGPT as an unexpected ally. The OpenAI model became his research assistant, guiding him through unfamiliar territory. ChatGPT suggested exploring immunotherapy and directed him to the Ramaciotti Center for Genomics at the University of New South Wales (UNSW), where he initiated a unique process. Personalized sequencing. At the center, Conyngham contacted Professor Martin Smith and paid $3,000 to sequence Rosie’s tumor DNA. Although they typically do not accept requests from individuals, Smith agreed after learning that Conyngham, as a data analyst, planned to interpret the results with ChatGPT's help. From sequencing to AlphaFold. With the genetic data, Conyngham used various AI tools, including Google DeepMind's AlphaFold, to analyze mutations and model the proteins involved in the tumor. He identified potential drugs and presented his findings to the UNSW team. The challenge of bureaucracy. Although researchers found a promising immunotherapeutic drug, the manufacturer refused to supply it for this use. Smith then proposed exploring mRNA vaccines. Conyngham agreed and, after preparing a 100-page report in two months, obtained the necessary ethical approval for the experiment. Vaccine in record time. Under the direction of Professor Pall Thordarson, UNSW produced the vaccine based on the formula generated by Conyngham. The only task left was administration, which was handled by Professor Rachel Allavena, an expert in canine immunotherapy at the University of Queensland. Conyngham traveled ten hours for Rosie to receive the first dose in December. Encouraging results. Researchers from UNSW and the University of Queensland confirmed that one of Rosie’s tumors was reduced by half. Allavena noted the improvement in the dog’s mood and appearance. Conyngham recounted how, six weeks after treatment, Rosie regained her vitality. Limitations and perspectives. Although the story is remarkable, it does not represent a definitive cure: only one tumor responded, and there are no controlled trials or long-term data. Conyngham acknowledges that it is not a miracle solution, but it has improved Rosie’s quality and life expectancy. The value of technical expertise. Conyngham’s experience in data science and machine learning was key to advancing the process. However, success also depended on collaboration with immunologists, RNA engineers, and veterinary oncologists. A replicable model? Professor Smith raised the question of why this approach is not applied to humans. The answer is that clinical trials require years and large investments, as well as robust evidence, which is lacking in this case. However, oncologist David Thomas is already exploring similar mRNA therapies in humans and highlights the potential of citizen science supported by AI. Towards a second vaccine. The process has demonstrated that AI can accelerate the transition from idea to experimental treatment. Thordarson emphasizes that Conyngham’s ability to generate an mRNA formula without biology training is an example of how AI democratizes research. UNSW is currently working on a second vaccine targeting the tumor that did not initially respond. Source: xataka.com