Open source · CC-BY 4.0

Open FHIR Genomics schema for cross-vendor NGS reports

The exact data contract UNMIRI's NGS Interpretation API uses to unify Foundation Medicine, Tempus, Caris, Guardant, Natera, and seven other vendor formats. Inspect it before you sign anything.

View on GitHubOr integrate the API instead

What's in the repo

  • • TypeScript and Python types for variant, biomarker, CDx flag, contraindication, and trial-match resources.
  • • Eight example payloads shaped from real vendor reports, not toy data.
  • • Vendor-format mapping tables for the 10+ vendors UNMIRI parses today.
  • • AMP/ASCO/CAP 2017 tier conventions: I-A, I-B, II-C, II-D, III, IV.
  • • evidence.externalLevels extension hook for licensed-implementation plug-ins (OncoKB, COSMIC, etc., under each implementation's own licensing terms).
  • • Identifier-only citation patterns: CIViC EIDs, ClinVar IDs, ClinicalTrials.gov NCT identifiers, PubMed IDs.

Why we open-sourced it

Verifiable contract

You can see the integration shape before you sign a contract. No NDA-gated PDFs, no "we'll send you the spec under MSA."

Working production schema

Most published FHIR Genomics examples come from spec authors. This one comes from the team parsing real vendor reports today.

Conventions matter most

Versioned parsers, identifier-only citations, AMP/ASCO/CAP as the public taxonomy. We made those decisions in the open and you can challenge them on GitHub.

How to use it

  1. Inspect: clone the repo, read examples/, read the mapping tables in vendors/.
  2. Integrate: import the TypeScript or Python types into your own pipeline and emit conformant Bundles.
  3. Contribute: open issues or PRs for missing fields, vendor edge cases, or convention questions.
  4. Buy: if you'd rather integrate UNMIRI's API instead of building parsers yourself, that's what we sell. /product/ngs-api.

License

CC-BY 4.0. Standard attribution: UNMIRI NGS-FHIR schema, https://github.com/unmirihealth/unmiri-ngs-fhir-schema. UNMIRI doesn't claim the schema is novel intellectual property — FHIR Genomics is the spec. The mapping tables, convention decisions, and working implementation are what we open-sourced.

Frequently asked questions

Why FHIR Genomics R4 instead of STU3?
R4 is the production-shipped variant of FHIR Genomics. UNMIRI's parser stack also supports STU3 Bundles for vendors that still emit them (see /coverage). The open schema standardizes on R4 because that's what FHIR Genomics implementers ship today.
Why not include OncoKB Levels?
OncoKB is a licensed proprietary knowledge base. Including its level taxonomy in an open schema would imply integration UNMIRI hasn't licensed and would constrain downstream implementers. The schema's evidence.externalLevels map is the documented hook for licensed implementers to plug in proprietary level systems under their own licensing terms.
Why not include COSMIC curated content?
Same posture. The cosmicIdfield is an identifier-only reference (license-clean), but COSMIC's curated annotations (FATHMM, recurrence counts, hot-spot context) require a commercial license and are not part of the open schema.
What about NCCN guidelines?
Not used and not paraphrased anywhere. NCCN content requires a commercial license. The schema cites AMP/ASCO/CAP 2017 as the public evidence-tier framework and cites primary sources directly (CIViC, ClinVar, ClinicalTrials.gov, openFDA, CPIC).
Can we use the schema in our commercial product?
Yes, under CC-BY 4.0. Standard attribution required. UNMIRI does not claim the schema is novel intellectual property — FHIR Genomics is the spec — but the working implementation, vendor-format mapping tables, and convention decisions are what we open-sourced.

Building on the schema?

Open issues on the repo. If you're evaluating whether to build vendor parsers in-house or buy the API, talk to us.

Open the repo →Talk to the team