AI FOR COMPARATIVE ONCOLOGY
We use AI and open biological data to search cancer across dogs, cats and humans for connections that might otherwise be missed.
Different species.
Shared biology.
More clues.
WHY THIS MATTERS
Cancer affects all of us. Looking across species can reveal patterns that are harder to see when we study people, dogs and cats separately.
Around four in ten people in the US will be diagnosed with cancer during their lifetime.
Rates and rankings vary by country, age and sex. Source: US National Cancer Institute ↗
Cancer is common in dogs, particularly as they get older. Around half of dogs over ten are affected by cancer.
Cancer is common in older cats, but high-quality population data are still surprisingly limited.
Feline population-level cancer data remain less complete than human and canine data.
CANCER, IN PLAIN ENGLISH
A human, a dog and a cat are obviously different. But cancer exploits some remarkably similar pieces of biology in all three.
Your body is made of cells. Normally they grow, divide and die under tight control. Cancer begins when cells escape those controls and keep growing when they shouldn't.
Dogs, cats and people naturally develop cancer. Some tumours use related genes and biological pathways, giving researchers three different views of the same underlying problem.
The evidence is spread across huge numbers of datasets, papers and experiments. AI can help connect those pieces and highlight patterns a researcher might not encounter in their own field.
That's the opportunity: look for what is shared, understand what is different, and find the experiment the existing evidence suggests should come next.
COMPARATIVE ONCOLOGY, IN PLAIN ENGLISH
Dogs, cats and people naturally develop many of the same cancers. The diseases aren't identical, but they can involve related genes, biological pathways and responses to treatment.
Comparative oncology studies those similarities and differences. Archie adds AI to help connect evidence scattered across thousands of datasets and research papers.
THE METHOD
From data to discovery, with evidence at every step.
Bring together cancer datasets and research from different species.
Look for genes, pathways and treatments that appear across species.
Search for something particularly valuable: the experiment nobody appears to have tried yet.
Nothing becomes a finding because AI said it. Every claim must link back to evidence and survive validation.
INSIDE ARCHIE
Archie doesn't simply ask an AI for cancer ideas. It starts with biological data, tests what changes, connects equivalent biology across species and keeps a record of every step.
ARCHIE-000C · BLIND VALIDATION
published biological terms independently recovered from the underlying canine lymphoma data.
Archie's result was cryptographically frozen before the published benchmark was introduced into the analysis environment.
CROSS-SPECIES BIOLOGY
Archie maps equivalent genes between species and asks whether cancer is disrupting related biological systems.
THE EVIDENCE GRAPH
Archie connects species, genes, biological pathways, datasets and published evidence. The goal is to make the gaps visible too.
MEET ARCHIE SCOUT
Archie Scout searches scientific databases for potentially useful cancer evidence. It can read at a scale no individual researcher reasonably can. But nothing Scout finds automatically becomes a fact.
FIRST MISSION · COMPLETE
What reusable molecular datasets and primary research are available for T-cell lymphoma in cats?
New evidence goes into quarantine first. Archie checks what the research actually says, preserves its provenance and tests the verification system itself. Only evidence that clears the appropriate checks can enter the public Knowledge Graph.
VERIFIER v0.2 · CONTROLLED EVALUATION
10 / 10Five relevant papers retained. Five deliberately irrelevant controls rejected. The test was repeated after the first version scored 9/10. No evidence was promoted automatically.
OPEN RESEARCH
Archie doesn't hide its work. Every completed research mission is designed to be understandable by people and interrogable by machines.
Browse the research questions, disagreements, original memos and source papers. Open the catalogue →
Read the investigation records, evidence status and original memo texts. Open the JSON catalogue →
Cancer research is too important for useful knowledge to remain trapped in papers, databases, institutions or AI conversations.
ARCHIE RESEARCH LAB
One AI searches. Another verifies. Another compares cancer across species. Another looks for experiments nobody appears to have done. Then another tries to prove them all wrong.
What question should we answer?
What evidence exists?
What does it actually say?
What biology is shared?
What's missing?
Can we prove the finding wrong?
What's the smallest useful experiment?
Can a normal person understand it?
A MISSING EXPERIMENT?
Archie checked whether its September PubMed search could identify a feline intervention study. The result describes that search and its limits.
A substantial cannabinoid and cancer research landscape exists, from laboratory biology to early clinical research.
Archie located experimental cannabinoid research in canine cancer, including lymphoma studies and work involving dogs receiving chemotherapy.
Archie has not yet located a direct published feline cannabinoid-cancer intervention experiment.
Scout ran targeted PubMed searches covering feline lymphoma, mammary carcinoma, squamous-cell carcinoma, sarcoma and other feline cancers against CBD, THC, cannabinoid receptors and the endocannabinoid system.
Only two candidate papers were returned. Neither represented the missing feline cannabinoid-cancer intervention experiment.
Archie is not claiming that cannabinoids cure cancer, or that nobody has ever performed such an experiment. It means Archie has not yet located a published experiment answering this particular question. Dataset and registry searches are still required.
What happens when feline cancer cells are exposed to carefully controlled cannabinoids such as CBD or THC?
THE BIOLOGY BEHIND THE QUESTION
Humans, dogs and cats have an endocannabinoid system: a biological signalling network that existed long before anyone discovered cannabis.
Our bodies naturally produce signalling molecules including and . Cannabis produces different molecules, including and , that can influence parts of this existing biological machinery.
That does not mean cannabis cures cancer. It means there is real biology worth investigating.
ARCHIE'S PLAIN-ENGLISH DICTIONARY
Cancer research is full of intimidating language. Archie translates the terminology without removing the science.
OPEN RESEARCH
A transparent record of what worked, what didn't and what we're investigating now.
Before asking Archie to find something new, we made it prove that its methods could recover biology already established by researchers.
The successful result was cryptographically frozen before the published benchmark was introduced. Statistical interpretation accounts for important limitations including the hierarchical, correlated nature of Gene Ontology terms.
See the full story →This investigation compares molecular evidence across canine, feline and human T-cell lymphoma to search for shared vulnerabilities and overlooked treatment opportunities.
Has one species already given us a cancer clue that researchers in another species haven't fully explored?
ARCHIE-000
Our first method found plausible cancer signals but failed its predefined validation test. We kept the result.
The second method found shared lymphoma biology but still failed the benchmark we had set. We kept that result too.
We redesigned the benchmark fairly, froze Archie's answer first, then revealed the published result. The overlap was substantial.
Science is not a slot machine where you keep pulling the handle until you get the answer you wanted. ARCHIE-000 exists to make that principle part of the software.
PROJECT UPDATES
Follow the published work from the corrected note back to the underlying records.
Read the human and canine evidence, the open question about cats and the citation corrections.
Read what each role wrote, including disagreements and the errors the audit later identified.
Scout ran 32 PubMed searches, returned two candidates and identified an unanswered question in that material.
THE DOG BEHIND THE NAME
Archie was a long-haired dachshund who died from lymphoma in September 2026.
His loss started a simple question: what if cancer research in dogs, cats and humans could learn more from each other?
The Archie Project is our attempt to find out.
A little dog. A big question.