Skip to content

DisMech assay and NAM inventory: issues and PRs by gingin77

Written 2026-10-13 against dismech main at commit 3eb0796817, from the GitHub history of monarch-initiative/dismech (first compiled 2026-10-08). Scope: issues and PRs authored, assigned or commented on by gingin77 whose focus is a method, assay, model system, NAM, or a framework for recording how a mechanism is measured. Disease curation work is left out. An issue and the PRs that answered it share one header. Status is as of the compile date. - The Adverse Outcome Pathway framework work (AOP and EMOD alignment, and seeding AOPs from dismech) is inventoried separately in the AOP framework inventory. - The schema elements a NAM touches are diagrammed in NAM schema elements, and a machine-readable version of this inventory is at data/nam-methods-inventory-2026-10-13.json. - The companion documents under data/ are deliberately not listed in the site navigation; this report is their entry point.

This report is aware of issue #13602, NAMs inventory - what fits and what's fringe, which sorts non-animal models by what ties the model to the disease. This report was not created to align with that analysis approach: it is organised by topic and by issue-PR pair, and it does not classify models by tie.

Contents

1. Organ-chips, organoids and iPSC models from the NAMeRS symposium

Epic: curate NAMs from the NAMeRS 2026 Summit (#4873, open, authored by mellybelly, commented)

The NAMeRS child PRs under this section were opened by repository agents, not by gingin77, and appear as context for the sub-issues split.

Opened to enrich entries with NAM evidence from the Triangle CERSI summit. Agent-authored child PRs added a droplet-microfluidic protease platform to Endometriosis, the Wyss lung organ-chip to NSCLC, three zebrafish anxiety assays to GAD, an iPSC liver-on-chip to acetaminophen hepatotoxicity, an EAC organ-chip, a bone-marrow-on-chip to chemotherapy-induced neutropenia, TNNT2 iPSC-cardiomyocytes to DCM, an iPSC sensory-neuron MPS to peripheral_axonal_degeneration, and Liver-Chip and organoid co-cultures to the DILI module. Blockers: hallucinated and misresolved PMIDs, an EAC chip modelling a different regimen than claimed, reference validation reporting zero checks on modules, and a scanner re-auditing the whole epic until gingin77 split out #9792, #9793 and #9794.

NSCLC lung organ-chip as a structured experimental model (#9792, closed; PR #9931, merged)

The Hassell 2017 lung organ-chip (PMID:29020635) was cited only as treatment evidence, so the model was invisible to the mechanism graph. PR #9665 added the ORGAN_ON_CHIP record; #9931 grounded its tyrosine kinase inhibitor readout in GO:0007169, keeping the relationship at MEASURES because the chip never claims to reproduce a driver lesion. Blockers: the paper is abstract-only, so quotable text ran out; gingin77 objected that the Tumor Microenvironment link overclaims since the chip has no immune or stromal compartment; the dormancy finding has no home node; two readouts remain ungrounded.

Esophageal squamous cell carcinoma patient-derived organoid (#9793, closed; PR #9797, merged)

ESCC had no organoid, chip or iPSC evidence while its adenocarcinoma sibling had a tumour-chip citation. The PR added a 24-patient ESCC organoid library (Nakagawa 2025, PMID:40169778) linked to a new Chemotherapy Resistance node with cisplatin, 5-FU and fedratinib readouts, and to TP53 Dysfunction, both RECAPITULATES at MODERATE fidelity. Blockers: no ESCC organ-chip literature exists, so an organoid stood in; a batched Europe PMC query returned the wrong abstract, forcing per-PMID fetches; the new node is isolated because the paper shows no upstream cause; organoids lack stroma and immune cells.

Scope decision for purely computational NAM evidence (#9794, open)

Filed because the in-silico strand of NAMeRS, virtual imaging trials, had no obvious home and needed a maintainer call. Nothing has been added and no decision is recorded. The triage bot argued the design decisions already exclude acquisition parameters and radiology decision support, which a virtual imaging trial simulates, and proposed judging by what a paper claims rather than its method, pointing instead at atrial fibrillation in-silico trials on the existing computational_models slot. COMPUTATIONAL evidence was thin when the issue was triaged on 2026-08-28 (642 items against 9,065 IN_VITRO) and remains so: at 3eb0796817, kb/disorders/ carries 1,136 COMPUTATIONAL against 20,943 IN_VITRO evidence_source lines, about 5 percent either way.

2. The NAMO rare-disease model bridge

Seven NAM models never reach the pathograph (#10432, closed; agent PR #10573; page fix PR #10430, merged)

Thirteen non-animal models were curated for the NAMO bridge, but seven in Stargardt, Noonan and Ulcerative Colitis had no modeled_mechanisms. #10573 wired retinal organoid and iPSC-RPE models to ABCA4 and RPE nodes, cortical organoid and iPSC-cardiomyocyte models to Noonan nodes, and colonoid, colon organoid and colon-on-chip models to barrier and immune nodes. #10430 made the project page link its five diseases. Blockers: the issue's candidate targets were wrong in three places, a RAF1 paper refuted the ERK to hypertrophy edge, the healthy-tissue colon-on-chip was capped at MEASURES and LOW, and no coverage metric exists for model wiring.

Tracking high-value NAM target diseases (#802, open, authored by cmungall, commented; PR #12582, merged)

The parking lot for diseases with strong organoid and MPS literature but no entry. gingin77 reported all seven now have entries and reframed it as a NAM-linkage checklist. #12582 linked models for six: BBSome retinitis pigmentosa animal models, Wilson disease iPSC hepatocytes and prime-edited liver organoids, LCA10 optic cups and QR-110-treated retinal organoids, TSC cortical spheroids and cerebral organoids with a FAILS_TO_RECAPITULATE link, a BRAF V600E colorectal PDO biobank, and an MSI-H organoid-lymphocyte co-culture. Blockers: the reviewer caught a mislabelled TSC limitation, opposite second-hit claims, an overstated LCA10 link; Wilson organoid fidelity stays UNKNOWN; Usher waits on #10822.

3. Liver fibrosis MPS and the fibrotic_response module

First liver fibrosis NAM on Liver_Cirrhosis and the DILI module (PR #9935, merged)

The three existing liver models were all drug-toxicity models and no fibrosis entry or module had an experimental model. Added the Akura Twin 384-well liver fibrosis MPS (PMID:40754287): HepaRG microtissues with THP-1 monocytes paired with stellate microtissues, challenged with TGF-beta1, methotrexate and acetaminophen, linked as CO_CULTURE to stellate activation, inflammation, hepatocyte injury and TGF-beta signalling. Blockers: no NAMO class fit until namo:CoCulture was found; MicrophysiologicalSystem requires microfluidics the plate lacks; monocyte activation was TGF-beta1-driven rather than drug-driven, so fidelity is LOW; the paper is abstract-only.

Experimental models for the fibrotic_response module (#9997, closed; PR #10282, merged)

The most-reused module had no experimental model, and the two organ-specific fibrosis NAMs were not fair stand-ins for fibrosis generally. #10282 added the organ-agnostic Scar-in-a-Jar macromolecular crowding fibroblast assay (PMID:34028741, with dermal and lung demonstrations) linked to Mesenchymal Cell Activation and Excessive ECM Deposition at MODERATE fidelity. Blockers: the fibrosis chip literature is entirely organ-partitioned; precision-cut slices were rejected as murine with no NAMO class; tissue_term had to be omitted; three CHANGES_REQUESTED rounds over readouts asserting direction on methods-only quotes; exogenous TGF-beta1 forces PERTURBS rather than RECAPITULATES.

Liver fibrosis NAM as the test case for the AOP Assay/NAM bridge (#10051, closed; PRs #10257 and #10299, merged)

The alignment page named Assay/NAM to ExperimentalModel as the cheapest bridge to AOP-Wiki and nothing had exercised it. #10257 mapped AOP 38, protein alkylation leading to liver fibrosis, onto the Akura Twin nodes: five of seven Key Events reach a node, both endpoints miss, and eight biological_scale tags were added. #10299 corrected the adverse outcome row, which had conflated fibrosis with cirrhosis. Blockers: nothing in the schema can reference a KE or KER and AOP-Wiki is not citable; the acetaminophen MIE link rests on a paywalled paper; biological_scale is single-valued; dismech has no organ-level liver fibrosis event.

4. The T-cell dependent antibody response (TDAR) assay

TDAR AOP project page (PR #11857), its accounting correction (PR #11905), and failed claims plus actions (PR #12003), all merged

Three AOP-Wiki pathways end at impaired TDAR, a standard immunotoxicology endpoint, and an OpenScientist run was commissioned to link TDAR to infection susceptibility. #11857 added projects/TDAR_AOP.md comparing the cluster against the germinal_center_reaction module. Findings: the TDAR adverse outcomes are terminal with no downstream KERs, the only quantitative bridge measures concanavalin A proliferation instead, Luster 1993 says no single immune test predicts host resistance, and both frameworks stop one node short of infection. #11905 corrected the page's understated list of unchecked sections; #12003 separated five failed claims from a twelve-finding action list. Blockers: the report's evidence streams skip the TDAR node; only a module conformance audit and one edge survived.

5. ToxCast and Tox21 assay endpoints

ToxCast as a structured reference source (#12682, open)

A chemical to assay hit-call is a citable fact nothing in the repository can verify, so a fabricated bisphenol A activity claim passes every gate. The issue specifies a StructuredSource emitting one cache file per chemical keyed on DTXSID with a TOXCAST: prefix, sized on the Jeong and Choi 2020 plastic-additive screen. Nothing has merged; the hit-call work survives only on an archived local branch, and the metadata half was split to #12858. Blockers: the EPA host fails DNS from sandboxed environments, the key must be requested by email, the sizing paper is paywalled, an API source does not fit the sha256 manifest pattern, and the assays slot has no validation path.

Map ToxCast assays to pathograph nodes using assay metadata (#12858, open; PRs #12923, #13226, #13228, #13236, merged)

ToxCast assays are molecular-level NAMs whose EPA CTX metadata could map an assay to a node independently of any chemical result. #12923 added just toxcast-refresh, a parser for 1,570 endpoints and the epa-ctx-api skill. #13226 added projects/TOXCAST.md and a coverage script with its report: 552 endpoints share a gene with 452 nodes in 317 entries. #13228 wrote the THRB TR-FRET coactivator endpoint against Generalized_Resistance_to_Thyroid_Hormone as a full model record that validates cleanly while asserting something false; #13236 pinned it in fixtures. Blockers: models live on Disease and links target same-file nodes, so an endpoint is copied everywhere with no shared identity; gene joins fail by lesion versus state, direction and quantity; cell-free assays have no model type; the API is unversioned.

GO-resolving biologicalProcessTarget values (#13224, open) and mitochondrial depolarization as a narrow bridge (#13225, open)

#13224 tested whether the 38 biologicalProcessTarget values bridge assay to node through GO. Nine resolve, covering 507 endpoints, but the top two are too broad to license an edge, leaving about 115 honest endpoints, with the H295R steroidogenesis assay (OECD TG 456) as the one substantive bridge. #13225 split out GO:0051882, six endpoints that are four measurements, reaching a PRKN-Related_Juvenile_Parkinson_Disease node and a Diethylene_Glycol_Poisoning node. No PR for either. Blockers: 29 values do not resolve, three carry EPA typos, H295R endpoints list the hormone's receptor as gene target, the HepG2 time series has no kinetic slot, and cmungall objected that GO cell death must not describe toxic cell death.

Endocrine Disruptor Screening Program history (#13161, closed; PR #13219, merged)

The estrogen signalling page claimed ESR1 is among the most heavily assayed ToxCast targets and the receptor the US statute was written around, and cited nothing. #13219 added a section with 17 verified identifiers covering the 1996 Food Quality Protection Act, OECD TG 455, 493 and 440 with the uterotrophic ring trial, the ToxCast ER pathway model EPA accepted in place of three EDSP Tier 1 assays, the androgen receptor model, and CERAPP and CoMPARA. Blockers: the 40-endpoint ESR1 figure is computed, not published; estrogen AOPs carry no OECD endorsement while androgen and thyroid AOPs do; AOP counts pin a date but not the CLI command.

6. Developmental neurotoxicity batteries and the zebrafish VAMR assay

DNT in vitro battery coverage of the pathograph (PR #13424, merged)

Opened to measure what the developmental neurotoxicity in vitro battery reaches, using the 17 processes in Tal et al. 2024. Added scripts/dnt_ivb_coverage.py behind just dnt-ivb-coverage, a report with 1,157 distinct links, and 14 tests. Findings: escape response, anxiety-like behaviour and epigenetic markers reach no node with a model attached; learning and memory reaches one in 242; neural network formation looks covered but most matched nodes describe hyperexcitability in epilepsies read out by EEG rather than a network being built; one organ-on-chip sits in the matched set. Blockers: this is lexical matching, not a mapping, so every count is an upper bound; three filter mistakes were documented; page anchors fail silently on renamed duplicates.

Zebrafish visual and acoustic motor response (VAMR) assay (#13450, open; PRs #13451 and #13452, merged)

Spath et al. 2026 (PMID:41780647) tested the VAMR behaviour assay on 17 DNT reference chemicals and dismech had no record of it. #13451 added the first animal_models entry on Organophosphate_Poisoning: chlorpyrifos in zebrafish larvae attached to Central cholinergic overstimulation as PERTURBS with PROXY_QUANTITY and SPECIES_MISMATCH divergences and four readouts. #13452 added a report, a TSV of all 884 concentration-response fits with a reproducible hit-rule script, and the finding that the paper's own hit rule contradicts its text in four places. Blockers: the printed Zenodo DOI does not resolve; behaviour only, no AChE readout; the scope gate forced the split; the reviewer overturned MEASURES; the open question is how to represent one assay once. Agent PR #13463 answered cmungall's ToxAssayMech request and found SOMAMech already overlaps (feasibility report).

7. Caenorhabditis elegans models for Major Depressive Disorder

Seed MDD animal models from C. elegans serotonin, insulin/FOXO and antidepressant pharmacology (#13749, closed; PRs #13755, #13756, #13757, #13759 and #13776, merged)

Major_Depressive_Disorder had twelve mechanism nodes and nine treatments but no model block. A scripted PubMed survey proposed worm work for the upstream half of the entry, never as a model of depression. Five PRs landed: the SSRI mechanism qualified with fluoxetine pharmacology showing reuptake inhibition is not a sufficient account; the tph-1 serotonin-null and daf-2 insulin-receptor immobility models as the first animal_models, at LOW fidelity with typed divergences; the risk-gene mutant panels recorded as a knowledge gap, not models; the serotonin-to-FOXO arm resolved as a HUMAN_MODEL_MISMATCH after an OpenScientist run; and the forced-swim validity dispute cited. Blockers: one-to-many worm-to-human orthology; the suicidality panels contain none of the entry's genes; the program rests on one laboratory and no divergence type expresses that; the FOXO chain holds only in the worm.

Mood disorders record no models at all (#13754, open)

None of the four mood-disorder entries recorded any model, yet together they already cite 14 model-organism papers inside their mechanism nodes, so the models exist but are invisible to queries and the pathograph. The issue lists what can be promoted without research, such as the GABA-A delta knockout mouse for postpartum depression and the Clock, GSK-3beta and Cacna1c models for bipolar disorder, and scopes deep research for hormone-withdrawal, mania, stress-paradigm and PMDD model families. Nothing has merged; the C. elegans work stayed on #13749. Blocker: most depression models are procedures on wild-type animals, and AnimalModel has no perturbation slot, which became #13768.

8. NAM scope, the lithium models, and the valproate and lithium report

NAMs inventory: what fits and what is fringe (#13602, open)

Filed because no rule says when a NAM belongs in a disease entry. It sorts by what ties the model to the disease: cells carrying the lesion fit, unaffected cells plus a causal exposure fit, unaffected cells plus a risk factor is undecided, unaffected cells plus the treatment is fringe, a wild-type target plus a chemical with no disease is fringe (the ToxCast case), and statistical enrichment is not a model link. Counts on main: 1,251 models in 600 entries, 212 links with no relationship. Nothing has merged. Blockers: no structured exposure, dose, duration or donor population, no field for the tie, one model nested in one file. Needs a maintainer decision.

Bipolar Disorder lithium-exposed neural progenitor and patient iPSC neuron models (#13600, open; PR #13788, merged; PR #13603, draft held)

Bipolar_Disorder had no experimental or animal models. The issue proposes lithium-exposed primary human cortical progenitors from 83 genotyped unaffected donors (Wolter 2023, Valone 2026) plus the already-cited Mertens 2015 patient iPSC neurons. #13788 landed the iPSC-derived dentate gyrus-like neurons as the entry's first patient-derived model: RECAPITULATES on Neuronal Hyperexcitability at MODERATE fidelity with a lithium-responder readout, MEASURES on Mitochondrial Dysfunction because the paper reports enhanced function. #13603 was narrowed to the progenitor model and stays held. Blockers: donors are unaffected, so it models lithium's action rather than the disease; no neurogenesis node; genotype-by-treatment effects have no field; the scope decision in #13602 is open.

Valproic acid and lithium gene-by-treatment report (#13787, merged; #13794, merged)

Opened to ask what one 2026 paper on valproate and lithium in genotyped human neural progenitors could add to dismech. The report names five entry gaps, four of which became work including #13600, and three schema limits: ExperimentalModel has no structured place for an exposure dose, a genotyped donor panel or a gene-by-exposure interaction, so the fetal valproate syndrome entry dropped the paper's headline response QTL result. #13794 corrected the claim that none were filed: the perturbation gap is #13768, the donor-panel gap sits in #13602, and the gene-by-exposure row has no issue and may belong in EnviroMech rather than dismech. Blockers: supplementary tables unread; two knowledge-base claims went stale between drafts.

9. Model organism, gene namespace and perturbation schema

AnimalModel cannot bind its organism or record a non-genetic perturbation (#13768, open; #13774, open)

Filed as the follow-on to #13548 after its organism and perturbation structures were set aside. AnimalModel.species is free text, written 66 ways across 2,381 models, and nothing structured can say what was done to a toxin-exposed zebrafish or a chronic-social-defeat mouse, so 351 procedure-based models bury the procedure in genotype, description or category. The proposal adds an NCBITaxon-bound organism and a perturbations list whose entries need no gene. #13774 asks for a report on the 50 whole-organism models still parked in experimental_models as OTHER. Nothing has merged. Blockers: species is an anchor key for entity references; migrating bound-organism records today would lose the binding; #13450 and #13754 are blocked on the slot.

Non-human gene identifiers on model genes (#13548, open, authored by caufieldjh, commented; #11140, open, authored by dragon-ai-agent, commented)

#13548 lets model genes be named in the model's own species through NCBIGene and UniProtKB, with an offline label gate. #11140 asked to document the older convention of recording the human ortholog. gingin77's comments, prompted by the worm survey in #13749, showed that convention is underdetermined for invertebrate and fish models: tph-1 has two human candidates, daf-2 three, daf-16 four, and two curators would pick differently. The namespace PR answers that by recording the worm gene as itself, with orthology moved to a planned structured link. Blockers: the PR awaits maintainer sign-off as a schema change; nine MGI bindings fall through as unchecked; a mouse dataset binding a mouse gene sits outside the gate's scope; documentation written against the old rule would be wrong within weeks.

10. Model fidelity and readout grading

Bgee cross-species model fidelity check (PR #10513, merged, authored by another contributor, commented)

A prototype script checks whether a model organism's ortholog of a disease gene is expressed in the anatomy where the human mechanism operates, a necessary but never sufficient condition for fidelity, with a proposal explaining why Bgee presence calls cannot be imported as evidence. gingin77's comment endorsed the check in toxicology terms as context of use: a NAM is chosen to match its context, and a model for advancing understanding of a complex disease differs from the minimal model for chemical safety screening. Blockers reported: 79 percent of links not evaluable from anatomy mismatch, 4 percent cell-type coverage, Ensembl returning empty ortholog lists for real genes.

Inverted readout grading on the Ganab mouse (PR #10350, merged)

Two evidence items on the Ganab heterozygous mouse's FAILS_TO_RECAPITULATE link in Polycystic_Kidney_Disease_3 were graded REFUTE where the quoted sentence is the null observation the readout reports. The PR regraded both to SUPPORT with DIRECT directness, since the negative sense is already carried by the relationship and the UNCHANGED direction. Not a new assay, but a rule for grading animal-model readouts. Blockers: nothing gates this error because check-snippet-grading deliberately does not gate supports; the reviewer found 14 occurrences across 12 other files; merge was blocked for days by a stale-base check.

A phenotype as the readout of a mechanism (#11427, open)

Curators wanting to say a phenotype is the readout of a mechanism were borrowing sequelae, which compiles to a phenotype-to-target edge and reverses causation; a census found nine such edges in six files. Nothing has merged. The bot showed the construct already exists as Phenotype.reports_on with READOUT_OF, CORRELATES_WITH, PREDICTS and PHARMACODYNAMIC_MARKER_OF, used in over 150 files, so this is a discoverability gap rather than a schema gap. The recommended fix is a reports_on migration and one sentence in CLAUDE.md.

Evidence strength, study design and assay typing (#9421, open, authored by another contributor, commented)

A generic evidence-model issue proposing study design, sample size and strength fields. gingin77's 2026-09-04 comment tied it to #10772: ECO types the evidence while OBI types the assay, 564 assay entries bind no OBI term and only two sit on a pathophysiology node, OBI has no adapter, and letting an evidence item point at its model would match EMOD's indirection, so one schema change should cover both issues. cmungall ruled that agents were redoing a hackathon discussion and that trial tiers do not suit pathograph internals built from converging animal, in vitro and computational lines. No change has merged.

Methods actually landed in kb/

Disorder entries and modules both live under kb/, and modules carry experimental models too, so this table covers both.

Method or assay Entry or module Item
Human orthotopic NSCLC lung organ-chip, TKI readout grounded in GO Non-Small_Cell_Lung_Cancer #9931 (record from #9665)
ESCC patient-derived organoid library Esophageal_Squamous_Cell_Carcinoma #9797
Akura Twin 384-well liver fibrosis MPS (co-culture) Liver_Cirrhosis, drug_induced_liver_injury #9935
Scar-in-a-Jar macromolecular crowding fibroblast assay fibrotic_response #10282
Retinal organoid, iPSC-RPE, cortical organoid, iPSC-cardiomyocytes, ALI colonoid, cytokine-conditioned colon organoid, colon-on-chip (links added) Stargardt_Disease, Noonan_Syndrome, Ulcerative_Colitis #10432 via #10573
ATP7B iPSC hepatocytes and prime-edited liver organoids; CEP290 optic cups and QR-110 retinal organoids; TSC2 cortical spheroids and cerebral organoids; BRAF V600E PDO/PDOX biobank; organoid-lymphocyte co-culture; BBSome RP animal models (links added) Wilsons_Disease, Leber_Congenital_Amaurosis_10, Tuberous_Sclerosis_Complex, BRAF_V600E_Mutant_Colorectal_Cancer, MSI_High_Colorectal_Cancer, BBSome-Related_Retinitis_Pigmentosa #12582
tph-1 serotonin-null and daf-2 insulin-receptor C. elegans animal models; SSRI mechanism qualified with worm fluoxetine pharmacology; serotonin-to-FOXO hypothesis with HUMAN_MODEL_MISMATCH record Major_Depressive_Disorder #13755, #13756, #13759, #13776
Patient iPSC-derived dentate gyrus-like neurons with lithium-responder readout Bipolar_Disorder #13788
Chlorpyrifos zebrafish VAMR behaviour animal model Organophosphate_Poisoning #13451

Tooling

Scripts, skills, reports and project pages that support assay and model curation but add nothing to kb/.

Tool Files Item
ToxCast/Tox21 endpoint annotation fetch and parser, just toxcast-refresh toxcast_assays.py #12923
epa-ctx-api skill for the EPA CompTox API SKILL.md #12923
ToxCast pathograph coverage script and report, just toxcast-coverage toxcast_pathograph_coverage.py, report, projects/TOXCAST.md #13226, #13228, #13236
DNT in vitro battery coverage script and report, just dnt-ivb-coverage dnt_ivb_coverage.py, report #13424
VAMR hit-call re-derivation script, TSV of 884 fits, and assay report vamr_endpoint_hit_calls.py, TSV, report #13452
Bgee cross-species model fidelity check (prototype, authored by another contributor; commented) bgee_model_fidelity.py, proposal #10513
Valproate and lithium gene-by-treatment fit report report #13787, #13794

Methods not incorporated into kb/

Assays and models that were surveyed, proposed or drafted but have no record in kb/, and why.

Method or assay Where it stands Item
Virtual imaging trials (Duke CVIT) and other purely computational NAMs Scope undecided; no maintainer disposition #9794
Tonsil explant and tumour-on-chip chemo-immunotherapy models from the NAMeRS summit Never curated; left on the epic #4873
Usher syndrome NAMs Waits on node structure #802, #10822
Murine precision-cut tissue slices across intestine, liver and kidney Recorded as a module discussion on organ divergence, not as a model; murine and no NAMO class #9997, #10313
ToxCast/Tox21 assay endpoints as records or evidence (1,570 endpoints; THRB TR-FRET worked example) Coverage and a worked example on the project page only; no endpoint name appears in kb/; schema has no disease-agnostic assay record #12858, #13228
ToxCast chemical hit-calls as a structured reference source Unmerged; parked on an archived local branch #12682
H295R steroidogenesis assay (OECD TG 456) via biologicalProcessTarget Proposal only; gene field names the receptor, not the enzyme #13224
Mitochondrial membrane potential endpoints (APR_HepG2_MitoMembPot, TOX21_MMP) via GO:0051882 Proposal only; endpoint identity has nowhere to live #13225
EDSP Tier 1 screens and the ToxCast ER and AR pathway models Cited on the estrogen project page as history, not as assays in kb/ #13219
DNT in vitro battery endpoints (17 processes) Lexical coverage report only; no crosswalk recorded in kb/ #13424
T-cell dependent antibody response (TDAR) assay AOP comparison on a project page; the assay is not recorded in kb/ #11857
VAMR assay for trichlorfon, ketamine and nicotine Later steps of the issue; open design question of recording one assay once #13450
Lithium-exposed primary human neural progenitors from 83 genotyped donors Draft held for the NAM scope decision #13600, #13603, #13602
C. elegans risk-gene mutant panels in a foraging-immobility assay Recorded as a knowledge gap, not as models; suicidality genes with no anchor in the entry #13757
GABA-A delta knockout mouse (postpartum depression); Clock, GSK-3beta and Cacna1c mice (bipolar); ESC/E(Z) lymphoblastoid cells (PMDD); stress paradigms and hormone-withdrawal models Already cited as evidence but not promoted to model records; stress and withdrawal paradigms blocked on the perturbation slot #13754, #13768
Procedure-based whole-organism models (toxin, diet, stress, surgery) generally No perturbation slot on AnimalModel; 50 such models parked in experimental_models as OTHER #13768, #13774

Blockers that recur across the inventory

  • No disease-agnostic assay record. experimental_models lives on Disease and a link must target a node in the same file, so a reusable endpoint is copied per disease with no shared identity (#12858, #13225, #13450, #13602).
  • The assays slot is bound to OBI with no OAK adapter or enum cache, and ECO is absent, so how a claim was measured cannot be recorded structurally (#10772, #11394, #12682, #9421).
  • Several of the primary papers behind the models curated here (the NSCLC organ-chip, the Akura Twin liver MPS, the ToxCast sizing paper and the zebrafish VAMR assay) were available only as abstracts or behind a paywall. Because dismech evidence must quote the cached reference, the quotable text in those cases ran out before the model's readouts were all covered (#9792, #9935, #12682, #13450).
  • Models involving co-cultures, gravity-linked plates and ex vivo slices, revealed some NAMO vocabulary gaps. Specifically, NAMO has no class for a non-perfused, gravity-linked microphysiological plate (its MPS class is abstract and requires microfluidics) or for ex vivo tissue slices, and the project page's in-scope list showed only four of NAMO's 56 classes, so one model was first left unbound for the wrong reason (#9935, #9997).
  • AnimalModel binds no organism and has no perturbation slot, so procedure-based and exposure-based whole-organism models have nowhere structured to say what was done to them (#13768, #13754, #13450, #13774).
  • Worm-to-human and fish-to-human gene orthology is one-to-many, and the human-ortholog convention on model genes forces a pick the biology does not support (#11140, #13548, #13749).
  • No structured exposure, dose, duration, kinetic series or donor population on a model (#13225, #13602, #13600, #13787).