Current normal-expression breadth

Descriptive policy oncoref.normal-expression-breadth.2026-11-04.1 uses bulk tiers from oncoref:tissue-hazards:v1:sha256:3370cb388d11e33820e0332bb224eac576cb1ff9bd7e7536db1cae6dd212cbda. No admission cutoff or membership change is introduced. The report covers 20,099 distinct member loci in the qualified reference universe, each once; these are gene profiles, not protein-group counts.

Skin, spleen and thymus now resolve to Protected. The historical October panel remains available as historical_panel(); anatomical_panel() uses the registry. The non-retinal eye residual uses the bulk eye rule. Cell exceptions never lower a whole-organ observation. Each organ has equal weight within its tier; brain regions share one organ's weight. Missing compartments retain their fixed weight.

For each tier, B = sum(w × x / (x + L)). Reproductive expression does not enter this calculation and cannot improve it. The separate reproductive concentration score mean(N / (N + R)) answers a different question.

Current descriptive baseline ceilings are {'vital': 4.0, 'protected': 8.0, 'manageable': 16.0} nTPM, without gene-specific IHC or clinical-credit adjustments. These anchors are research choices, not injury thresholds. The before/after table separates historical tiers/ceilings, current tiers with historical ceilings, and current tiers/ceilings. All scores include brain-weight sensitivity 0.5, 1 and 2. Panels expose every weight and assessment ID.

The current and historical profiles retain native source, version, unit, measurement and missing coverage. Lower/upper bounds describe missing anatomical weight, not confidence intervals. An entirely missing profile remains [0, 1]. HPA consensus is a published aggregate, not an independent donor cohort. Duplicate observations cannot increase weight.

Donor summaries and donor burdens remain separate. The available four-donor CTAG2-heart series supports a median but not the prespecified ten-donor q90; missing q90 is not zero. Marginal bootstrap intervals do not provide a joint tissue interval or population-risk estimate.

Rebuild with python scripts/build_current_normal_breadth.py. The receipt pins source content, ontology, ceilings, code and outputs. The October report stays frozen and reproducible.