How it works
Deterministic first. AI explains second. Auditable throughout.
The authority is a deterministic engine: the same input always returns the same result, traceable to its evidence. A plain-language AI layer sits on top for people who aren't geneticists — but it never overrides the engine.
Two layers, in order
The deterministic engine is the source of truth. Given a genotype, it produces a reproducible interpretation grounded in CPIC (clinical pharmacogenetics guidelines) and PharmVar (variant nomenclature). The AI layer then translates that result into readable language. The engine decides; the AI never does.
The pipeline
- Genotype in. A patient or cohort genotype, resolved against the five 1000 Genomes superpopulations — 2,504 real phase-3 samples.
- Deterministic engine. The authoritative, auditable interpretation, grounded in CPIC and PharmVar.
- Auditable output. A trial-ready result with provenance and a call status, plus a plain-language explanation.
Three honest call statuses
Every interpretation resolves to one of:
called— evidence sufficient; a result is returned.cannot_call— the engine refuses rather than guess, and cites the rule.insufficient_data— not enough signal to decide, stated plainly.
A recommendation you can’t defend is worse than no recommendation
When the evidence behind a call isn’t strong enough, the engine refuses rather than guesses — and every refusal names the specific rule that triggered it. That is what makes an output defensible to a reviewer. The Evidence Sufficiency Layer holds 12 decision rules, 10 verdict rules, 9 uncertainty rules, and 3 bias-detection kinds.
Results come stratified across five superpopulations plus a global mix, and are exported as trial-ready rows for downstream workflows.