[iskandr]

Summary

• Add Manageable risk between Disposable and High consequence; retain Reproductive and Vital.
• Narrow Disposable to cutaneous melanocytes, follicles, tonsil, appendix and gallbladder. Skin/spleen/thymus/thyroid/adipose are Manageable risk; salivary glands and cartilage are High consequence.
• Implement the human HPA v23 category-by-IHC RNA grid (missing/uncertain, Approved-negative, Supported-negative, Enhanced-negative): Vital 1/2/3/6; High consequence 2/4/6/12; Manageable risk 5/10/15/30 nTPM.
• Preserve the independent veto for any positive IHC outside permitted tissues, including Low. Relaxation requires complete same-organ routine negative IHC and uses the weakest reliability.
• Add literature citations and explicit distinctions between anatomy, injury consequences and local control. Export five-category human/dog/mouse inventories without transferring human tolerability or numeric thresholds.
• Release 1.10.0; heavy reference DATA_VERSION remains unchanged.

Report results

All 2,537 candidates were reevaluated from checksum-pinned original HPA observations:

Scenario	Retained
Previous adaptive	654
New anatomy, previous RNA limits	629
New anatomy, fixed baseline	582
New anatomy, adaptive grid	624

1 gained (TEX15), 31 lost. CT83 is excluded at salivary RNA 2.6 nTPM with unavailable IHC (baseline 2). PRAME, NY-ESO-1, MAGE-A3/A4 and SSX2 remain retained. CTAG2 passes expression gates but retains its existing manual holdback. Exact causes and the two stages of comparison are exported for every candidate.

Verification

• Naming follow-up: all candidate decisions and 126,262 per-observation RNA/IHC audit rows match after renaming; readable labels are included in mapping exports. The existing protected API key displays as High consequence.
• 123 focused tests passed across taxonomy, gene filters, adaptive ceilings, independent IHC gate and species expression.
• Ruff lint/format passed.
• Strict MkDocs build passed; comparison figure visually checked.
• All 17 generated report outputs and implementation hashes verified. Historical audit artifacts are unchanged.

Interpretation

The numerical grid is an explicit project screening policy, not a clinically validated safety limit or a GTEx-to-HPA conversion. Manageable risk still excludes positive IHC. Cartilage, whole eye and inner ear lack routine bulk measurements in the archived human panel; a category assignment does not establish absence. No mouse expression atlas or maternal/fetal negative-coverage panel is invented.