Synthetic data only. Educational use only. UNMIRI is a developer preview, not a medical device, and is not for clinical decision-making.
Pathogenic BRCA1 in high-grade serous ovarian
The starting input
A Caris MI Profile sample report. High-grade serous ovarian, FFPE biopsy. The Caris layout interleaves IHC with NGS, so the parser has to keep section context straight. Sample at data/sample_reports/caris/caris_mi-profile_breast.pdf (templated for ovarian, watermarked "Synthetic data, demonstration only").
What the parser extracts
- One variant: BRCA1 c.181T>G p.Cys61Gly, classified pathogenic in the report header
- HRD score: positive (Genomic Instability Score 65)
- IHC: ER positive, PR positive, HER2 0
- MSI: stable
- TMB: 6 mut/Mb (not high in this tumor type)
What the evidence join adds
- CIViC + ClinVar: BRCA1 Cys61Gly is a well-characterized pathogenic founder allele
- openFDA: olaparib and niraparib indications include BRCA1-mutant or HRD-positive ovarian maintenance
- AMP/ASCO/CAP tier: I-A (FDA CDx + biomarker on guideline-equivalent footing)
What the CDS surface produces
The PARP-inhibitor branch fires only when both the variant call and the HRD biomarker are positive. Engine 2 expresses this as a deterministic combined predicate, not a model. The recommendations card surfaces the maintenance-line PARP option (olaparib first, niraparib as alternative), platinum-sensitivity context, and a trials block filtered to ovarian HRD-positive cohorts. Educational-use banner stays sticky.
What surprises buyers
The HRD score is a derived biomarker. Caris reports it directly, but Tempus and Foundation Medicine compute and label it differently. UNMIRI normalizes all three into the same biomarker entry with provenance pointing back to the vendor's text. That makes "HRD-positive" a portable concept across the three vendors' reports without giving up the receipts.