Pilot programme open to a limited number of clinical genetics teams.Apply for a place

The investigation workbenchfor unsolved genetic cases

Many rare-disease cases stay open after exome or genome testing. The evidence is there. It is spread over too many places.

Evidence is spread outThe variant list, the clinic notes, the pedigree and the papers sit in different tools.
The same lookups, againNothing records which genes were already checked, or why they were set aside.
Reasoning stays in headsWhen the person who ruled a gene out moves on, the reason leaves with them.
Clinical geneticistsGenetic counsellorsVariant scientistsBioinformaticiansLab leadsDepartment heads
For the people who work on cases that stayed open.

A case record that does the arithmetic and shows its sources. It ranks, cites and re-checks. A person decides.

ExplainedEvery rank links to the record that produced it. A score is never a probability.
A person decidesA clinician confirms suggestions. A person signs classifications and reports.
VersionedPipeline, knowledge and model versions are recorded with every result.
AI models

Point to where to look, and say why.

Some answers sit in a gene with no disease link yet. The models rank established and unestablished candidates on separate lists and give reasons you can check at each step.

Reads the patient, not just the variants.

Compares the patient’s phenotype with what is known about genes and diseases and ranks the genes whose profile fits best. A feature that is unknown or not tested is never held against a gene.

Focal seizurespresent
Ataxiaunknown
Retinal findingsnot tested
Discovery

Candidate new genes, kept apart.

Genes with no established disease link are ranked on their own list. Each is labelled, explained and never mixed with established findings.

Each suggestion gives its reasons
Reasons cite curated sources
Reviewed and signed like any other lead
Built for reviewThe models decide where to look and say why. Your team reviews the evidence and makes every clinical decision.
Versioned and traceableEach result records the knowledge and model release behind it. When knowledge changes, the case shows what changed.
Tested, with limits statedHow the models were evaluated, and where they fall short, is written up on the validation page. Validation
What goes in
VCF files
Aligned reads (BAM or CRAM)
Raw reads (FASTQ)
Phenotype, pedigree and clinic notes
Single patients or whole families
One case record, whichever file arrives.
Intake

Intake that does not repeat itself.

Referral, history, phenotype, family, consents and files are entered once. Clear statements in the notes become provisional terms. Uncertain ones wait for you.

Guided case creation
Notes read into terms
Pedigree and consents recorded once
Discovery

Hypotheses and next steps.

Genes are ranked from independent channels, with what supports each one, what argues against it and what is still unknown. Then the next test that would change the picture.

Ranks with their reasons
Next steps name the ACMG criterion they would add
What-if simulation
Assistant

A case assistant, and re-checks.

Ask about one case or widen the question to your other cases. Each statement cites its source and anything unverified is marked. When knowledge changes, affected open cases are checked again.

Cited answers
Assessment drafts to review
Open cases re-checked on new knowledge
Review

Classify, review, sign.

ACMG/AMP criteria are proposed with their evidence, and a person confirms and signs. Reports are rendered from the case record, so they match the evidence.

ACMG/AMP criteria with sources
Tasks, notes and sign-off
Reports rendered from the record
Knowledge

Knowledge versions

Each new release re-checks the open cases it touches.

October 2026
New gene–disease links
One open case re-checked, one new lead.
Late September 2026
Population frequencies
Rarity thresholds re-evaluated.
September 2026
Phenotype ontology
Terms remapped, affected terms flagged.
Pipeline · August 2026
Pipeline update
Every run keeps the pipeline version it used.
Example timeline from a synthetic case. Names and dates are illustrative.
Decision support. Never a diagnosis, never without a person’s sign-off.

Start with one casethat stayed open.