Presentation
Which targets does a tumor actually display?
We model which peptide fragments a specific tumor naturally presents on its surface — not just what could bind in theory, but what is really there for the immune system to see.
AI-native oncology research
Finding the true targets the immune system can see.
A tumor hides in plain sight. Veritope builds AI to identify the specific targets a patient’s own immune system can recognize on their cancer — then to act on them. We start where we can learn fastest: companion-animal oncology, built to translate to human cancer.
veri · tope
An epitope is the exact fragment the immune system recognizes. A veritope is the true one — the target that is really there, and really seen.
Most candidate targets are neither.
The problem
Most “cancer targets” are predictions, not sightings.
The field is full of candidate targets that bind in a tube but are never actually displayed on a real tumor — or that a T cell can’t tell apart from healthy tissue. Chasing those costs years and patients.
The hard, valuable question isn’t “what might work?” It’s “what is this specific tumor truly presenting, and what can the immune system truly recognize?” Answering it well is a data and modeling problem — the kind disciplined AI is actually built for.
- ✕ Binding in a tube ≠ displayed on a tumor
- ✕ Displayed ≠ recognized by a T cell
- ✕ Recognized ≠ safe against healthy tissue
- ✓ Veritope keeps these claims separate — and measures each one
The approach
Three engines, kept honest and kept separate.
We refuse to collapse cancer into one magic score. Each engine answers a different question, is measured against the strongest available baselines, and is validated on data it has never seen.
Recognition
Which of those can a T cell truly recognize?
A displayed target is only useful if an immune receptor can engage it — selectively, without mistaking it for healthy tissue. We predict recognition, not just binding.
Design
When nature falls short, can we engineer recognition?
Where a patient lacks a natural response, computational protein design offers an orthogonal route — building recognition molecules against a verified target.
How we work
We separate what we know from what we hope.
Credibility in this field comes from one habit: never letting a hypothesis wear the clothes of a result. Everything we publish is tagged by how sure we are — and we try hardest to prove ourselves wrong.
What we know
Published, reproducible evidence — the ground we stand on. We cite it and we re-check it.
What we believe
Hypotheses we find promising but unproven. We label them as such, and design tests that could falsify them.
What we’ll prove
The specific, pre-registered result that would move a belief into knowledge — measured against strong baselines, on data the model has never seen.
Why pets first
Companion animals get cancer like we do — and deserve better options.
Dogs develop spontaneous tumors with real biological parallels to human disease. Working in veterinary oncology lets us build and validate our platform with genuine ground truth, shorter timelines, and families who want every good option for their animal.
As we find approaches that work, the same engines are designed to translate toward human cancer — carefully, and only when the evidence earns it.
Go deeper
Read our research brief.
The full, candid account of what we’re attempting, what the biology requires, and how we plan to prove it — written for scientists, and for the collaborators and partners we want to work with.