Specimen document — illustrative
Genomarker Diligence · case file
DC-2026-041 · Series A target — biomarker claims
Review complete4 computational claims examinedengaged by: life-science fund (confidential)DELIVERABLES SIGNED
3 OF 4 SUPPORTED
The core biomarker claim (CL-01) is methodologically defensible, independently replayed, and preserved at Grade A. The headline ML performance claim (CL-02) is not supported: feature selection was performed before the train/test split, and independent replay with the leak removed yields AUC 0.68. Findings, evidence packages, and replay attestations for every claim accompany this report.
| Claim | Analysis · profile | Assurance | Replay | Preservation | Verdict |
|---|---|---|---|---|---|
| CL-01 · Biomarker panel separates responders | DE · GM-BIOMARKER-DE-1.3 | Pass | REPRODUCED_EQUIVALENT | Grade A | Supported as stated |
| CL-02 · Model predicts response, AUC 0.91 | ML · GM-BIOMARKER-ML-1.0 | Failed | DIVERGED | Grade B | Not supported — leakage |
| CL-03 · Pathway enrichment in treated arm | Enrichment · exploratory | Disclosed | REPRODUCED_EXACT | Grade A | Supported · exploratory only |
| CL-04 · Effect replicates in public cohort | DE · GM-EXPLORATORY-DE-1 | Pass | REPRODUCED_EQUIVALENT | Grade C | Supported · dependency referenced |
Deliverables — every claim, independently inspectable
- Assurance statements4 signed
- Structured findings23 findings
- Verdict reportthis document
- Evidence packages.gmk × 4
- Replay attestations4 signed
- Preservation assessmentgraded
The receiving party does not have to trust this report. Every deliverable verifies offline with genomarker verify — the evidence, not the reviewer, carries the claim.