Missing has_part relations to CL: systematic scan of term definitions (33 evidence-backed candidates)

Following the pattern of #2904 and #2960, this is a proposal list of has_part relations from Uberon structures to CL cell types that are asserted in a term's own textual definition but missing as an axiom.

No ontology changes are being proposed in a PR here — this ticket is the report, so the list can be reviewed (or cherry-picked) before anything is added.

How the candidates were found

1. Parsed all 15,621 non-obsolete [Term] stanzas in src/ontology/uberon-edit.obo.
2. Indexed every non-obsolete CL label and synonym from src/ontology/imports/local-cl.owl (3,219 classes, 5,413 label/synonym strings).
3. Matched multi-word CL names inside each term's textual definition, discarding pairs where the has_part already exists.
   → 113 candidate subjects.
4. Triaged against the anti-patterns described in #2963 and against the 247 has_part CL:* axioms already in the edit file.
5. Found one reference per assertion and verified every PMID against Europe PMC.

Proposed axioms (33)

Each would carry a {source="PMID:…"} qualifier, matching the evidence-backed sub-pattern already used for e.g. islet of Langerhans and olfactory epithelium.

Retina

• UBERON:0001791 inner nuclear layer of retina → CL:0000636 Mueller cell — PMID:16839797
• UBERON:0001792 ganglionic layer of retina → CL:0000740 retinal ganglion cell — PMID:23083731

Cerebral cortex (matches the existing cortical layer III / cortical layer V axioms)

• UBERON:0005391 cortical layer II → CL:0000598 pyramidal neuron — PMID:25622573
• UBERON:0005395 cortical layer VI → CL:0000598 pyramidal neuron — PMID:25622573

Hippocampus

• UBERON:0002313 hippocampus pyramidal layer → CL:1001571 hippocampal pyramidal neuron — PMID:2687721
• UBERON:0014548 pyramidal layer of CA1 → CL:1001571 hippocampal pyramidal neuron — PMID:2687721
• UBERON:0014549 pyramidal layer of CA2 → CL:1001571 hippocampal pyramidal neuron — PMID:2687721
• UBERON:0014550 pyramidal layer of CA3 → CL:1001571 hippocampal pyramidal neuron — PMID:2687721
• UBERON:0002191 subiculum → CL:0000598 pyramidal neuron — PMID:2687721

subiculum uses the generic pyramidal neuron rather than hippocampal pyramidal neuron, since the subiculum is part of the hippocampal formation but not of the hippocampus proper.

Basal ganglia

• UBERON:0002435 striatum → CL:1001474 medium spiny neuron — PMID:21469956
• UBERON:0001882 nucleus accumbens → CL:1001474 medium spiny neuron — PMID:21469956

Olfactory bulb

• UBERON:0004186 olfactory bulb mitral cell layer → CL:1001502 mitral cell — PMID:25232305
• UBERON:0002264 olfactory bulb → CL:1001502 mitral cell — PMID:25232305
• UBERON:0002264 olfactory bulb → CL:1001503 olfactory bulb tufted cell — PMID:25232305
• UBERON:0002264 olfactory bulb → CL:1001435 periglomerular cell — PMID:25232305

Brainstem and PNS

• UBERON:0001719 nucleus ambiguus → CL:0000100 motor neuron — PMID:3667964
• UBERON:0000044 dorsal root ganglion → CL:1001451 sensory neuron of dorsal root ganglion — PMID:25420068
• UBERON:0001700 geniculate ganglion → CL:0000101 sensory neuron — PMID:28970527
• UBERON:0005362 vagus X ganglion → CL:0000101 sensory neuron — PMID:31116992
• UBERON:0000395 cochlear ganglion → CL:0011113 spiral ganglion neuron — PMID:30209249
• UBERON:0004670 ependyma → CL:0000065 ependymal cell — PMID:20024659

Asserting on vagus X ganglion rather than separately on inferior vagus X ganglion, and on ependyma rather than on brain ependyma / spinal cord ependyma, since those is_a children inherit the axiom.

Kidney

• UBERON:0000074 renal glomerulus → CL:0000653 podocyte — PMID:25918223
• UBERON:0000074 renal glomerulus → CL:1000742 glomerular mesangial cell — PMID:25918223
• UBERON:0000074 renal glomerulus → CL:0002188 glomerular endothelial cell — PMID:25918223

Digestive tract

• UBERON:0001902 epithelium of small intestine → CL:0000584 enterocyte — PMID:24326621
• UBERON:0001902 epithelium of small intestine → CL:0000160 goblet cell — PMID:24326621
• UBERON:0013743 base of crypt of Lieberkuhn of small intestine → CL:0000510 paneth cell — PMID:23398152
• UBERON:0013743 base of crypt of Lieberkuhn of small intestine → CL:0009017 intestinal crypt stem cell of small intestine — PMID:24326621

Inner ear

• UBERON:0002227 spiral organ of cochlea → CL:0000589 cochlear inner hair cell — PMID:28915323
• UBERON:0002227 spiral organ of cochlea → CL:0000601 cochlear outer hair cell — PMID:28915323
• UBERON:0002227 spiral organ of cochlea → CL:0002163 internal pillar cell of cochlea — PMID:23545368
• UBERON:0002227 spiral organ of cochlea → CL:0002164 external pillar cell of cochlea — PMID:23545368

Lymphatic system

• UBERON:0034953 embryonic lymph sac → CL:0002138 endothelial cell of lymphatic vessel — PMID:22859612

References cited

PMID	Citation
PMID:2687721	Amaral & Witter (1989) The three-dimensional organization of the hippocampal formation: a review of anatomical data. Neuroscience
PMID:3667964	Bieger & Hopkins (1987) Viscerotopic representation of the upper alimentary tract in the medulla oblongata in the rat: the nucleus ambiguus. J Comp Neurol
PMID:16839797	Bringmann et al. (2006) Müller cells in the healthy and diseased retina. Prog Retin Eye Res
PMID:20024659	Del Bigio (2010) Ependymal cells: biology and pathology. Acta Neuropathol
PMID:21469956	Gerfen & Surmeier (2011) Modulation of striatal projection systems by dopamine. Annu Rev Neurosci
PMID:22859612	Yang et al. (2012) Lymphatic endothelial progenitors bud from the cardinal vein and intersomitic vessels in mammalian embryos. Blood
PMID:23083731	Masland (2012) The neuronal organization of the retina. Neuron
PMID:23398152	Clevers & Bevins (2013) Paneth cells: maestros of the small intestinal crypts. Annu Rev Physiol
PMID:23545368	Wan, Corfas & Stone (2013) Inner ear supporting cells: rethinking the silent majority. Semin Cell Dev Biol
PMID:24326621	Barker (2014) Adult intestinal stem cells: critical drivers of epithelial homeostasis and regeneration. Nat Rev Mol Cell Biol
PMID:25232305	Nagayama, Homma & Imamura (2014) Neuronal organization of olfactory bulb circuits. Front Neural Circuits
PMID:25420068	Usoskin et al. (2015) Unbiased classification of sensory neuron types by large-scale single-cell RNA sequencing. Nat Neurosci
PMID:25622573	Harris & Shepherd (2015) The neocortical circuit: themes and variations. Nat Neurosci
PMID:25918223	Scott & Quaggin (2015) Review series: The cell biology of renal filtration. J Cell Biol
PMID:28915323	Fettiplace (2017) Hair cell transduction, tuning, and synaptic transmission in the mammalian cochlea. Compr Physiol
PMID:28970527	Dvoryanchikov et al. (2017) Transcriptomes and neurotransmitter profiles of classes of gustatory and somatosensory neurons in the geniculate ganglion. Nat Commun
PMID:30209249	Petitpré et al. (2018) Neuronal heterogeneity and stereotyped connectivity in the auditory afferent system. Nat Commun
PMID:31116992	Kupari et al. (2019) An atlas of vagal sensory neurons and their molecular specialization. Cell Rep

Candidates deliberately rejected

Roughly 80 of the 113 were discarded. Grouped by reason:

• Broad structure + specific cell type (the primary anti-pattern in #2963): ganglion, cranial ganglion, ganglion of peripheral nervous system, neural nucleus, gray matter, gray matter of spinal cord, cell group, cerebellar layer, cerebellar cortex.
• The match is a cell part, not a whole cell, so has_part would be false — axon tracts, dendritic fields and synaptic structures: cranial nerve II, optic tract, nerve fiber, white matter of spinal cord, brain white matter, white matter of cerebellum, arbor vitae, the three parallel fiber terms (granule-cell axons), skeletal muscle fiber triad, cerebellar glomerulus, hippocampus stratum oriens / stratum radiatum / stratum lucidum and the CA1–CA3 stratum terms, dentate gyrus molecular layer, olfactory bulb external plexiform layer, olfactory bulb glomerular layer, Merkel nerve ending (see #1803), nociceptor nerve ending, sublaminar layer S5.
• Structure surrounds rather than contains the cell: endomysium and its smooth/cardiac/skeletal variants, glomerular basement membrane.
• Already covered: Purkinje cell layer of cerebellar cortex and dentate gyrus granule cell layer (composed_primarily_of); granular layer of cerebellar cortex, photoreceptor layer of retina, cortical layer III, cortical layer V (a more specific has_part is already asserted).
• Deliberate alternative modelling: outer nuclear layer of retina uses overlaps CL:0000287 {notes="soma"} because only the soma lies within the layer. Left alone.
• Text-match false positives: kidney → erythrocyte, tectorial membrane of cochlea → trichogen cell (an insect term), rhombic lip → sensory epithelial cell, skeletal muscle organ, vertebrate (filler absent from the CL import).

Plausible, but needs a curator decision

Subject	Suggested filler	Open question
glomerular epithelium (+ metanephric / mesonephric variants)	podocyte	The term also subsumes the parietal (Bowman's capsule) epithelium, which has no podocytes. glomerular visceral epithelium already carries the axiom.
inferior glossopharyngeal IX ganglion	sensory neuron	Correct, but no petrosal-specific reference was found that could be verified.
cortical layer IV	spiny stellate neuron	No CL class for "spiny stellate cell"; stellate neuron (CL:0000122) is probably too broad.
epithelium of small intestine	paneth cell, tuft cell	Paneth cells are absent in several mammals (cat, dog, pig), so the assertion is only proposed at the small-intestine crypt base. A taxon GCI would be the alternative.
dentate gyrus subgranular zone	neural stem cell / granule cell progenitor	Needs a filler choice.
hilus of dentate gyrus	mossy cell	No CL class for the hilar mossy cell.
parotid gland striated duct	basal cell	Filler is very generic.
ovarian cortex	primary oocyte	Stage-dependent; may belong on the follicle instead.
insect eo-type sensillum	trichogen, thecogen, tormogen cell	Well-supported by the definition, but wants an insect-anatomy reviewer.
insect proneural cluster, insect trichome, testicular sheath, Kupffer's vesicle / yolk syncytial layer group	various	ZFA/FBbt-derived; better handled by the respective community curators.

has_part to other anatomical structures (UBERON → UBERON)

The same scan restricted to compositional phrasing ("composed of", "consists of", "contains", "made up of", "formed by") plus a UBERON label match yields 788 subjects / 1,359 pairs. These are not proposed, for two reasons:

1. docs/uberon-editor-sop.md states that for anatomy-to-anatomy links "preference should be given to using part_of rather than has_part" — most of these are better expressed (and usually already are) as an inverse part_of on the part.
2. Signal-to-noise is poor: matches such as "peripheral nervous system → central nervous system", "mouth → digestive tract" and "tube → anatomical structure" come from definitions that mention the other term without asserting composition.

A per-system pass, in the style of #2904 / #2960, is the workable way to approach that set. Happy to triage a system at a time if useful.

Implementation

The 33 axioms above are implemented and validated on the branch claude/uberon-missing-has-part-jlis59 (2 commits, +33 lines in uberon-edit.obo, no deletions), but no PR has been opened pending review of this list. On that branch:

• every PMID was verified against Europe PMC;
• robot convert round-trip is byte-stable, so the roundtrip_obo QC target passes;
• ELK reasoning over uberon-edit.obo plus its imports completes with no unsatisfiable classes;
• nine of the referenced CL classes are not yet in imports/merged_import.owl, so ROBOT wrote those axioms without the trailing ! label comment — the labels return after the next make refresh-imports, since the import seed is derived from the edit file.

Say the word and I will open the PR, or trim it to whichever subset survives review.

@ai4c-agent

Start by reviewing src/ontology/uberon-edit.obo, docs/uberon-editor-sop.md, and the 33 candidate axioms listed in the issue. Compare the implemented claude/uberon-missing-has-part-jlis59 branch with the proposal and check that each assertion has the cited PMID source qualifier. Done means the reviewed axioms are validated and ready for a pull request.