secretory cell and capable of some endocrine hormone secretion

[raymond91125]

CL term
CL_0000163 endocrine cell

Suggested revision of logical definition
A proposal has been made to merge GO:0060986 endocrine hormone secretion into GO:0046879 hormone secretion.
Please see geneontology/go-ontology#17084.

[dosumis]

There's an element of buck passing here. No-one wants to take charge of defining terms like endocrine. GO just passed the buck to CL & Uberon who were formerly relying on GO to define it.

The bigger picture:

https://en.wikipedia.org/wiki/Endocrine_system

The elements of the endocrine system can only be defined functionally - it consists of organs and cells that secrete hormones into the circulatory system through which they travel to stimulate distant cells/organs.

The same molecule may act in an endocrine, paracrine and autocrine fashion (or even as a neurotransmitter or neuromodulator) in different contexts. Serotonin is a good example:

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5526216/

In the past we've tried to deal with this by using role terms to extend the meaning of the chemical entity class (CHEBI has these) but that has never really worked. I think the most straightforward solution is to come up with ways to define that context by extending the definition of the process. e.g.

endocrine cell
EquivalentTo: cell that 'capable of' some ('hormone secretion and has target end location some 'circulating fluid'*
subClassOf: part_of some 'endocrine system'

'pancreatic b cell' subClassOf capable of 'insulin secretion' and has target end location](http://purl.obolibrary.org/obo/RO_0002339) some 'blood'
=> 'pancreatic b cell' subClassOf 'endocrine cell' (subClassOf part_of some 'endocrine system')

*Uberon TBA - parent to blood and hemolymph - this means the solution will work for arthropods too. If we want to restrict to vertebrates, we could just use 'blood'

My preferred solution would be for GO to define these process terms, as I think this is better for co-ordination + organismal terms. I also think that this general approach makes sense for GO. They really should have separate process terms for, for example, serotonin secretion in the gut and serotonin secretion as neurotransmitter given that there are likely to be differences in mechanism and regulation + the medical importance of the distinction. However, I suspect that they will be resistant to this.

How to proceed?

I will make the suggestion on the GO ticket. However, I think best not to bother waiting long for a response. To make this approach work, we need a mini project dedicated to it. This should:

(a) Extend the relations guide to document this pattern. (I think using subClassOf axioms is going to be more practical in most cases. It is not impossible to use DOSDPs for this but would require some thought and infrastructure work so we should avoid making this a dependency).
(b) work through various endocrine cells and organs in CL & Uberon implementing and checking inference. (Fixing for one general class is pointless)
If GO does end up creating the process terms, we could always update.

Missing piece:
Co-ordination across ontologies is always a challenge. Ideally we'd have some rigorous ontology design pattern system that is shared across interdependent ontologies. However we have neither the time nor the resources for this. Instead I think we need some centralised, human readable documentation for major shared design patterns. This mini project could be a proof-of-concept pilot for this approach.

Ontologies that could benefit from shared patterns: CL, Uberon, FBbt. Less directly: GO & phenotype ontologies (who might want to refer to disruption of specific endocrine secretory processes)

[shawntanzk]

Solution from @dosumis

I think the most straightforward solution is to come up with ways to define that context by extending the definition of the process. e.g.

endocrine cell
EquivalentTo: cell that 'capable of' some ('hormone secretion and has target end location some 'circulating fluid'*
subClassOf: part_of some 'endocrine system'

'pancreatic b cell' subClassOf capable of 'insulin secretion' and has target end location](http://purl.obolibrary.org/obo/RO_0002339) some 'blood'
=> 'pancreatic b cell' subClassOf 'endocrine cell' (subClassOf part_of some 'endocrine system')

or

For GO to define it (I'm guessing this ticket geneontology/go-ontology#24473)

(First one would be faster, but I think given that GO has a PR open already, perhaps it might be wise to just wait for that and do changes then)

[ghost]

@dosumis, will you clarify, is the next step to change the EA of 'endocrine cell' as follows:

CURRENT:

'secretory cell'
 and ('capable of' some 'endocrine hormone secretion')

REVISED:

cell 
and 'capable of' some ('hormone secretion' and ('has target end location' some 'haemolymphatic fluid'))

[dosumis]

GO have gone with UBERON:0001009 ! circulatory system so lets stick with that.

We then need a way to define (inferable) subclasses.

I suggest we go with

capable_of some (%s and ('has target end location' some 'circulatory system'))

Where %s is filled by some subClassOf GO 'hormone secretion'

e.g.
This term,

rdfs:label "type D enteroendocrine cell"
definition "A cell found throughout the gastrointestinal tract and in the pancreas. They secrete somatostatin in both an endocrine and paracrine manner..."

Should have

subClassOf capable_of some (somatostatin secretion and ('has target end location' some 'circulatory system'))

[dosumis]

CC @aleixpuigb

[gouttegd]

One comment about the proposed pattern:

Constructs like the following:

'capable of' some ('hormone secretion' and ('has target end location' some 'circulatory system'))

cannot be represented in the OBO Flat File format with standard OBO tags. Instead, they end up in the owl-axioms header in OWL Functional Syntax.

Now, I don’t think we should avoid such constructs just because they are “untranslatable” in OBO. I am mentioning that just so that editors of ontologies that use OBO as their edit format (e.g. Uberon, FBbt) are aware of the issue.

[dosumis]

They could live in fbbt-ext.owl (component).

[dosumis]

This ticket had now been superseded by this #1884 & subtickets.