Matching DNA variants to medicines takes scarce experts.
Cogenics turns variant-to-drug interpretation into a query, grounded in published prescribing guidelines.
Cogenics maps a patient's DNA variants to how they will respond to a medicine — grounded in CPIC prescribing guidelines and ClinVar, and built to run as a small specialized model.
Why it exists
Cogenics turns variant-to-drug interpretation into a query, grounded in published prescribing guidelines.
A Lancet study of 7,000 patients found it cut adverse drug reactions by about 30%. Cogenics is built to make that guidance available beyond specialist centres.
CPIC prescribing guidelines define what a correct answer is; ClinVar is the reference variant dataset. Every variant-to-drug association is validated against these sources before it becomes training material.
Following the PGxAI Andromeda validation template: blinded expert review plus guideline-adherence scoring, with gene-drug pair coverage tracked alongside accuracy.
What it does
Maps DNA variants to medicine compatibility and expected response.
Answers trace back to CPIC prescribing guidelines, not to model intuition.
Variant-to-drug associations are checked against ClinVar before training.
A specialized model that serves through Ember on hardware you control.
Process
Variant-to-drug associations are checked against CPIC and ClinVar.
A small specialized model is trained on the validated associations only.
Blinded expert review and guideline-adherence scoring, published with stated limits.
Access
Cogenics is validating its variant-to-drug associations before training. Follow it for validation results and research partnerships.
Follow CogenicsPharmacogenomics model
Questions / answers
Cogenics maps patient DNA variants to medicine compatibility, grounded in CPIC prescribing guidelines and ClinVar.
Every association is validated against published guidelines before training, and the model is measured by blinded expert review.
No. Cogenics is a research tool, not a clinical product, and makes no diagnostic or treatment claims.
Models are served through Ember, on hardware we control.