Variant analysis and interpretation
AI-powered Variant Interpretation Software with ACMG/AMP Classification
Varion assembles the evidence for every candidate variant on a GA4GH VRS store, applies ACMG/AMP scoring, and prioritizes the cases that matter, then hands each one to an analyst to weigh and sign. The software is customized to a lab's panels, runs on-premise, and never issues a report without expert approval.
Human-in-the-loop throughout. The analyst makes the call.
Variant interpretation is the real bottleneck in genomic testing
Once sequencing is fast, interpretation becomes the slowest stage of the test. An analyst queries ClinVar, gnomAD, and OMIM for every candidate variant and applies ACMG/AMP criteria by hand, and annotation API costs compound because nothing is cached locally. Most of that time goes to gathering evidence rather than to the judgment the test actually depends on.
What Varion includes
Varion arrives as a working interpretation engine that a lab shapes to its own panels and evidence sources.
- An ACMG/AMP rule engine that applies the criteria consistently across every analyst
- Parallel scoring tracks that combine rule-based classification, phenotype similarity, and AI pathogenicity scoring
- Automatic evidence assembly from ClinVar, gnomAD, OMIM, and related sources
- A local annotation cache that removes the per-query API cost at scale
- A GA4GH VRS-compliant variant store, with FHIR and HL7 delivery downstream
- Human-in-the-loop sign-off, so no report is issued without analyst approval
What's ready to run, and what the lab tailors
The rule engine, the evidence assembly, and the variant store already exist, so a project starts around the halfway mark rather than at zero.
| Already built | Customized for the lab |
|---|---|
| ACMG/AMP engine, scoring tracks, evidence assembly, VRS store, annotation cache | The gene panels and evidence sources, the classification rules and thresholds, the sign-off workflow and report routing |
Because that foundation is in place, NonStop estimates a build reaches production 40 to 50 percent faster than starting from scratch, and the lab owns the result when the engagement ends.
Between the pipeline and the report, with the analyst in control
Varion takes pipeline output, assembles the evidence, and hands the analyst a prioritized case. The classification and the sign-off stay with the qualified expert, and approved results flow downstream to reporting.
| Receives | Variant calls from the bioinformatics pipeline |
| Builds on | A shared GA4GH VRS store with the knowledge base (GENVAR) |
| Feeds | Approved results to reporting over FHIR and HL7 (ReportStudio) |
| Deploys | On-premise or private cloud |
From hours a case, down to under an hour
| Manual interpretation | With Varion |
|---|---|
| 6–8 hrs a case for classification and prioritization | 30–60 min a case, consistent across analysts |
* Figures are NonStop estimates and vary by engagement scope.
The aim is to return the hours lost to database queries, so an analyst's attention goes to the variants that genuinely need human judgment, which is where the value of the test lies.
Frequently Asked Questions
Does Varion replace the analyst's ACMG/AMP classification?
How does it reduce annotation API costs?
Is the variant store standards-based?
Is Varion an off-the-shelf product?
Can it run on-premise?
Book a call
See Varion on your own cases
Book a 15-minute scoping call and tell us the lab's panels and evidence sources. We will show where evidence assembly fits and what the analysts would still control.
Book a call →