The dog cancer vaccine pipeline is real — here is every tool, every step, and what it actually costs

A viral case study reveals how researcher Paul Conyngham designed a personalized mRNA cancer vaccine for his dog using a seven-step AI pipeline. The process began with commercial DNA sequencing of tumor tissue (~$3,000), followed by using ChatGPT as a research collaborator to interpret mutation data and iterate on vaccine design. The core of the pipeline leveraged Google DeepMind's open-source AlphaFold for protein structure prediction and free, open-source bioinformatics tools like pVACtools and NetMHCpan for the machine learning task of selecting the best vaccine targets (neoantigens). The total computational cost for these AI-driven steps was remarkably low, under $100 in cloud credits.

The final output was a half-page mRNA sequence specification, which was then synthesized by a university lab (UNSW RNA Institute) in under two months. The entire design phase was dwarfed by the three-month wait for ethics approval and administration, highlighting a non-scientific bottleneck. The case demonstrates that the technical pipeline for personalized cancer vaccines is now accessible, combining off-the-shelf AI models with established bioinformatics. Furthermore, the field is advancing rapidly; Isomorphic Labs recently released IsoDDE, a model reportedly 2x more accurate than AlphaFold 3 for this specific task. This project underscores a critical question: if such a pipeline works for a dog, the primary barrier to human rollout is regulatory, not scientific.