Quantitative genetics, fully auditable.
The science is deterministic, explainable, and auditable. Working results stay separate from released, versioned facts, so every genetic claim is governed and reproducible.
Live today
The decision support most lab software doesn't have.
These capabilities run today. They operate on released, versioned facts, never on raw or unreviewed values.
Recessive-risk pairing
Punnett-square offspring outcomes (affected, carrier, or clear) computed from released carrier statuses, not guessed.
Inbreeding coefficient (COI)
A pedigree-based inbreeding coefficient for a proposed pairing, with the common ancestors driving it traced for you.
Mate recommendation
Ranks candidate mates by a balance of safety (recessive risk) and diversity (COI), with hard filters to exclude unsafe pairings.
Governed normalization
Raw results map to canonical interpretations through inspectable definitions; unknown values fail closed into review queues.
Verifiable reports
Released results become immutable snapshots with verification metadata, so a report can be independently proven authentic.
QC & panel coverage
Submissions carry coverage and quality metrics; QC flags and coverage checks gate what is allowed to reach release.
On the roadmap
Parentage & relationship inference.
Built against the same governed model. The scientific design is set, with production rollout following validation against known trios and per-species cutoffs.
Parentage verification
Genotype-based parentage using Mendelian exclusion, with a likelihood-based confidence measure rather than a bare yes or no.
Relationship inference
Genotype-based kinship estimation that places pairs of animals in a degree of relationship and distinguishes close relationships from one another.
Identity & mix-up detection
Unusually high similarity between samples flags a likely sample mix-up or duplicate record before it contaminates downstream facts.
The honest boundary
AI assists. The engine decides. Nothing is released autonomously.
AI and automation help triage, surface candidates, and speed review. They do not assert genetic facts. Every released result passes through explicit review, approval, and release states recorded in the audit trail.
That separation is the point: the deterministic engine and versioned rules produce the science, a human approves the release, and the report carries a verifiable snapshot. No model silently changes published truth.
Rule governance
Every rule change is reviewed before it touches a result.
Interpretation rules are grouped into versioned rule sets. A change moves through a defined lifecycle, and you can see its impact before it goes live, so the science evolves without surprises.
Draft → review → Active
Rules are editable in Draft, locked for review, then activated. Activating a version automatically deprecates the previous one.
Fixtures
Curated genotype inputs with expected outcomes must pass before a rule set can be promoted, so a change can't ship if it breaks known cases.
Impact analysis
Before activation, preview exactly how many released results would change, and which would be added or removed.
Shadow mode
Run a new rule set in parallel against real submissions and log the differences, with no effect on released results, until you trust it.
Why it holds up
Governed truth, not report-only truth.
Facts are stored separately from reports, versioned, and reproducible. When rules or panels improve, results are reprocessed and re-reviewed without rewriting history.
Snapshot on release
Released results snapshot the rule and explanation-template versions, so later edits never alter an already-published report.
Reproducible
Every released fact ties back to the rule version and reviewer that produced it, so any result can be traced and re-derived.
Missing data is first-class
A referenced locus with no fact yields a structured "NotTested" outcome that shows in reports, never a silently skipped rule.
Put the science to work on your animals.
If you run breeding decisions or a testing lab, we'd like to show you the engine on real data.