unify naming conventions and taxon axioms for species-specific cell types

[cmungall]

CL is a frankenstein ontology when it comes to taxon specificity:

1. historic sensus, to genera and above; e.g CL:0000662 ! neuroglioblast (sensu Nematoda)
2. sensus to 3 letter codes with expresses axioms; e.g: CL:0000754 ! type 2 cone bipolar cell (sensu Mus)
3. 4 letter code with expresses axioms; e.g. CL:4023007 ! L2/3 bipolar vip GABAergic cortical interneuron (Mmus)
4. common names with has plasma membrane part axioms; e.g. CL:0002426 ! CD11b-positive, CD27-positive natural killer cell, mouse

The axiomatization is also pretty uneven - here are sample edges:

subject	subject_label	predicate	predicate_label	object	object_label
CL:0000754	type 2 cone bipolar cell (sensu Mus)	RO:0002292	expresses	PR:P50481	LIM/homeobox protein Lhx3 (mouse)
CL:0000754	type 2 cone bipolar cell (sensu Mus)	RO:0002292	expresses	PR:Q8R4I7	neuropilin and tolloid-like protein 1 (mouse)
CL:0000754	type 2 cone bipolar cell (sensu Mus)	RO:0002292	expresses	PR:Q9ER75	iroquois-class homeodomain protein IRX-6 (mouse)
CL:0002426	CD11b-positive, CD27-positive natural killer cell, mouse	RO:0002104	has plasma membrane part	PR:000001012	integrin alpha-M
CL:0002426	CD11b-positive, CD27-positive natural killer cell, mouse	RO:0002104	has plasma membrane part	PR:000001963	CD27 molecule
CL:4023007	L2/3 bipolar vip GABAergic cortical interneuron (Mmus)	RO:0002292	expresses	PR:P32648	VIP peptides (mouse)

• Sometimes a species-level PRO term is used, sometimes gene level, is there a principle here?
• I get the expresses vs plasma membrane relations reflect different experimental modalities for the cell types but this still feels unsatisfying and frankenstein-like

Also note for these 2, there is no direct OR inferred taxon constraints:

CL:0000662 ! neuroglioblast (sensu Nematoda) CL:0000754 ! type 2 cone bipolar cell (sensu Mus)

Recommendations:

• drop the string "sensu"
• use PRO conventions for taxon name consistently (human, mouse, worm, etc)
• always use species-specific PROs when the cell type is species specific
• consistently populate in-taxon axioms

[lubianat]

I agree with @cmungall comments; I'd say it is also a good opportunity to standardize which taxa are in the scope of CL.

PRO is clear on species-neutrality x species-specificity; without taxonomic genera-level entries. It is always neutral or "organism". If we default to PRO conventions, we should avoid having "mouse" meaning both the genus Mus or the species Mus musculus.

Maybe CL should provide just species-neutral terms for metazoa and species-specific terms for Homo sapiens and Mus musculus, adopting PRO's "(human)" and "(mouse)" nomenclature. This seems to be generally the scope of CL already, it is just not specified anywhere.

Genera-and-above terms could be created on a case-by-case basis and have the taxonomic name under parenthesis, dropping the "sensu" word.

How hard would it be to have species-specific terms should be generated automatically by some pipeline parsing taxon constraints? This could be a way to enforce consistency.

[cmungall]

Like Uberon, the scope of CL is all metazoa (excepting the most general classes that can be reused across all of life) The focus is vertebrates, with particular emphasis on human and mouse. Even more so than Uberon, because the mouse ssAOs don't have cell types (well, MA, has two cell types, and EMAPA 10 cell types, which is odd).

Great idea to have an automated check for the naming conventions.

[lubianat]

Related issue:

• https://github.com/obophenotype/cell-ontology/issues/1829

[github-actions[bot]]

This issue has not seen any activity in the past 6 months; it will be closed automatically in one year from now if no action is taken.

[cmungall]

let's not close this