Can DisMech and NAMO hold what an OECD Test Guideline says?
An OECD Test Guideline describes validated test methods in free text. For that content to be citable rather than paraphrased, a method's identity, the biology it observes, and its regulatory standing all have to land in a slot. This table checks one Test Guideline against the two schemas available, asking for each fact the document states whether DisMech has a slot, whether NAMO has a slot, and how good the fit is in each.
Companion to the schema elements a NAM touches, which maps the
DisMech side in detail, and to the NAMO vocabulary gaps met while curating fibrosis
NAMs, which records the same kind of finding from the
model-system direction rather than the regulatory one. Machine-readable form in
oecd-tg442d-schema-coverage-2026-11-02.json.
Written 2026-11-02.
Source document. OECD Test No. 442D, In Vitro Skin Sensitisation: Assays addressing the Adverse Outcome Pathway Key Event on Keratinocyte activation, adopted 2 July 2026 (10.1787/9789264229822-en). It describes five test methods across four appendices: KeratinoSens™, LuSens, EpiSensA, Epi2SensA and α-Sens®. The facts below come from the General Introduction and from each appendix's Initial Considerations and Principle of the Test sections.
Schema versions. DisMech at aef740d (2026-10-31); NAMO at 0a743c5 (2026-08-01).
Legend
| Symbol | Meaning |
|---|---|
| ✅ | A slot exists whose range and documented purpose fit this fact |
| ⚠️ | A slot exists but the fit is partial — wrong granularity, free text where structure is wanted, or a slot documented for another use |
| ❌ | No slot; the fact can only be carried as prose, or not at all |
Coverage
| TG 442D fact | DisMech slot | NAMO slot | DisMech | NAMO |
|---|---|---|---|---|
| Method name | name |
NamedThing.name |
✅ | ✅ |
| Cell line and its derivation | cell_types, cell_source |
CellularSystem.cell_types / cell_source; CellLineModel.passage_range, authentication_method |
✅ | ✅ |
| Model category (2D line vs reconstructed human epidermis) | experimental_model_type enum |
class hierarchy (CellLineModel vs ThreeDCellCulture) plus biological_organization_level, complexity_level |
⚠️ no RhE value | ✅ |
| Tissue model product (LabCyte EPI-MODEL24, EpiDerm™ EPI-200) | culture_system (free text), tissue_term (UBERON) |
three_d_architecture, matrix_composition |
⚠️ free text | ⚠️ no product-identity slot |
| Sensing pathway (Keap1 → Nrf2 → ARE) | readout biological_processes (GO-bound) |
Pathway (pathway_database, pathway_id) |
✅ | ✅ |
| Marker genes (ATF3, IL-8, GCLM, DNAJB4) | ExperimentalReadout.biomarker_term (NCIT) |
Gene (gene_symbol, ensembl_id, fold_change, p_value) |
⚠️ NCIT, not HGNC; no fold-change | ✅ |
| Which mechanism the method measures | modeled_mechanisms[].target → a pathophysiology node |
ModelsRelationship.biological_system_modeled |
✅ | ⚠️ not a pathograph node |
| Limitations (no oxidative metabolism, hence the pro-hapten blind spot) | limitations, typed divergences |
StructuredConcordanceResult quantifies match but carries no typed caveat |
✅ | ⚠️ |
| Inter- and intra-laboratory reproducibility (~85%) | — | Reproducibility |
❌ | ✅ |
| Performance against the animal reference (accuracy 77%, sensitivity 78%, specificity 76% vs LLNA) | — | ModelPerformance (accuracy, sensitivity, specificity, auc) |
❌ | ✅ |
| Numeric cut-offs (EC1.5; ATF3 15-fold, GCLM 2-fold) | — | QualityControlMetric.threshold, FunctionalAssay.reference_value |
❌ | ⚠️ repurposable |
| Viability constraint (≥80% EpiSensA, ≥60% Epi2SensA; IC30, IC50) | — | as above | ❌ | ⚠️ repurposable |
| Readout direction | ExperimentalReadout.direction enum |
FunctionalAssay.assay_result, Gene.fold_change |
✅ | ⚠️ no direction enum |
| Test Guideline number and appendix (442D, Appendix IA) | — | Reference only (id, title, year, url) |
❌ | ❌ |
| AOP and key event addressed (AOP 40, KE2 keratinocyte activation) | — | — (AOPs appear in schema prose only) | ❌ | ❌ |
| Reporter construct (luciferase under the ARE of human AKR1C2; Firefly Luc+ with Renilla-TK) | — | — | ❌ | ❌ |
| Validation route (validated reference method vs Performance-Standards-based) | — | — | ❌ | ❌ |
| Approving body and its role (EURL ECVAM / ESAC; JaCVAM; ICATM cooperation) | — | — | ❌ | ❌ |
| Regulatory use (UN GHS Category 1; within an IATA; TG 497 defined approaches) | — | Study.context_of_use, free text |
❌ | ⚠️ |
| Mutual Acceptance of Data status | — | — | ❌ | ❌ |
| Proficiency substances | — | — | ❌ | ❌ |
| Methods cited but not described (TG 406, 429, 442A, 442B) | — | — | ❌ | ❌ |
What the two columns say together
The schemas are complementary, not redundant. NAMO holds the quantitative facts
DisMech has no slot for — ModelPerformance, Reproducibility — and models a marker gene
as a gene, with fold-change and significance beside it, where DisMech can only name it as
an NCIT biomarker. DisMech holds the mechanism linkage and the typed caveats NAMO has no
slot for: a model links to a named pathophysiology node, and its shortfalls are recorded
as classified divergences rather than as prose.
Neither holds a method's regulatory standing. Eight rows are ❌ in both columns, and seven of them belong to one group: the Test Guideline identity, the AOP key event, the validation route, the approving body, Mutual Acceptance of Data, proficiency substances and cited comparators. That group is what makes a method regulatorily usable rather than merely scientifically described, and it is what a Test Guideline exists to establish. Nothing in either schema asserts it today. The eighth row absent in both, the reporter construct, is a biology fact rather than a regulatory one.
Two DisMech gaps are narrow and separately fixable. The model-type enum has no value
for reconstructed human epidermis, so EpiSensA and Epi2SensA would both be typed OTHER.
And a DisMech readout can name a gene only as an NCIT biomarker — BiomarkerTerm admits
gene products and imposes no hierarchy — not as an HGNC gene the way gene_term does
elsewhere in the schema, and with nowhere to put the fold-induction value or its
threshold. That matters for the two methods TG 442D describes as quantifying gene
expression:
The test methods currently described in this Test Guideline include three in vitro ARE-Nrf2 luciferase test methods and two test methods based on gene expression quantification
The two are EpiSensA and Epi2SensA, which read endogenous ATF3, IL-8, GCLM and DNAJB4 by RT-qPCR. The other three report an ARE-driven luciferase reporter, which is an engineered construct rather than an endogenous transcript, and so is a different curation problem.
NAMO has OECD only as a prefix. OECD: is declared and used in two example CURIEs.
There is no class for a Test Guideline, no enum for acceptance or validation status, and no
slot for an approving body.
Caveats on reading this
- One Test Guideline, chosen as a worked case. TG 442C and 442E are built the same way, but a Test Guideline from another endpoint area may carry facts this table does not anticipate.
⚠️ repurposablein the NAMO column means a slot's range fits but its documented purpose names a different use —QualityControlMetric.thresholdstanding in for an EC1.5 cut-off, for instance. Those are a stretch rather than a fit.- The table records what the schemas can hold. It says nothing about which repository should hold it, which is a separate and still-open question.