The Science: How We Prepare a Personalized Design

What RosieVaccine prepares for review

Only outputs the pipeline generates today.

An annotated target-ranking report — the candidate targets found for this dog, ranked, with the evidence behind each one.

A record of the quality and safety checks that ran — which checks were applied to this case and what they returned. Results are recorded for expert review, not used to decide anything automatically.

Sequence files in standard formats — the construct design, exported in the formats a manufacturer works from.

Model names, thresholds and internal formats stay inside the package, for the people reviewing it.

Evidence and responsibilities

What the evidence shows, where the limits are, and how to read it. Who does what is on the For Vets page.

Where predictive models help — and where humans decide

Predictive models help prioritize possible cancer-specific targets. Reproducible software applies quality and safety checks. Qualified people remain responsible for scientific, clinical, and regulatory decisions.

Predictive models Prioritize candidate targets Score biological patterns to rank the changes most likely to matter for this dog's cancer. A score is not proof a target works in a dog.

Reproducible software Applies quality & safety checks The same inputs produce the same design, with every step recorded in the evidence trail. Automation doesn't remove the need for human review.

Qualified people decide Whether a case proceeds at any step. Every diagnosis and treatment decision (veterinarian). Whether a design is scientifically sound (expert review). What may be manufactured or administered (regulators).

The human precedent

Personalized mRNA oncology has now met its endpoints in a phase 3 human trial. In August 2026, an individualized mRNA neoantigen therapy met phase 3 endpoints in resected human melanoma — recurrence-free survival and distant metastasis-free survival. No effect size has been published for that trial yet. In the same programme’s randomised phase 2b (KEYNOTE-942, n=157), the therapy added to pembrolizumab reduced the risk of recurrence or death by ~49% at five years (HR 0.51, 95% CI 0.294–0.887). In resected pancreatic cancer, a personalized vaccine induced neoantigen-specific T-cells in about half of a 16-patient phase 1 cohort.

This is scientific precedent for the method — not proof of efficacy in dogs. Species, tumors, and immune genetics differ.

What is known in dogs

Published canine evidence A conditionally licensed canine melanoma DNA vaccine: longer survival vs historical controls in its licensure study (n=58+53); a later single-centre retrospective (n=45) found no improvement. Mixed — and labelled as such.

Human precedent A phase 3 human trial of a personalized mRNA neoantigen therapy met its endpoints in August 2026; the randomised phase 2b (KEYNOTE-942, n=157) reported HR 0.51 at five years. Precedent for the method — not proof in dogs.

Still being validated No canine efficacy data exists for this exact approach. Our validation work is ongoing and unpublished.

Where we are today, in plain terms → About RosieVaccine

Citations

PMID 38246194Weber et al., Lancet 2024 — KEYNOTE-942, randomised phase 2b, resected human melanoma, mRNA-4157 + pembrolizumab, n=157.

PMID 37165196Rojas et al., Nature 2023 — autogene cevumeran, phase 1, human pancreatic ductal adenocarcinoma, n=16.

PMID 22126691Grosenbaugh et al., AJVR 2011 — huTyr DNA vaccine licensure study, canine oral melanoma, prospective vs historical controls, n=58+53.

PMID 23909996Ottnod et al., Vet Comp Oncol 2013 — retrospective single-centre review, canine oral melanoma ± Oncept, n=45.

UNSW 2025UNSW Newsroom — Paul Conyngham and Rosie: personal account; n=1, multimodal care, not affiliated with RosieVaccine.

USDA-APHISCenter for Veterinary Biologics — jurisdiction and guidance for veterinary biologics.