Need clarification of partonomy in modelling traits and phenotypes

[rays22]

Partonomy in modelling traits and phenotypes

I would like to understand how OBA models traits with regards to partonomy so that we can be consistent when adding new trait patterns.

Some of the GCIs suggest that the partonomic modelling of traits in OBA may be different from how phenotypes are modelled in uPheno.
For example, see GCI in entity_attribute.yaml

# (Q and characteristic_of some part-of some W) SubClassOf part-of some (Q and characteristic_of-in some W)
# https://github.com/obophenotype/bio-attribute-ontology/issues/8
logical_axioms:
  - axiom_type: GCI
    text: "%s and ('characteristic_of' some ('part_of' some %s)) SubClassOf 'part_of' some (%s and 'characteristic_of' some %s)"
    vars:
      - attribute
      - entity
      - attribute
      - entity

How to interpret this GCI, when one trait does not look intuitively part of another, even though there is some partonomic relationship between the bearers?

It makes sense for like a 'duration of M phase' to be part of an 'duration of cell cycle' instance. The 'mass of femur' instance can be part of a 'mass of hindlimb' instance.

However, what if the qualities of the parts are not additive?

For example, the acidity of a 'stomach neuroendocrine cell' intuitively does not seem to be part of the 'acidity of my stomach' even if the cell itself is part of my stomach.
Is the 'shape of my tonsil' is part of the 'shape of my head'?

---

Related issue: #139 (comment)

[dosumis]

Looks unsafe to me. @cmungall - What's the rationale. Can it be deleted or modified to make safer?

[matentzn]

I suggest to remove these for now - they are not doing any work right now - and make a new issue to consider adding them back.

[cmungall]

If they are not doing any work something has gone wrong.

See #8 (one of the first issues in this repo) for more background

I think it would be a massive mistake for OBA to switch to is-a overloading as is common in phenotype ontologies, and the dreadful subq pattern.

The way this ontology was set up originally followed TO (plant trait ontology), with no is-a overloading, and no assertion of is-as.

I do agree that the current rules are too inclusive, e.g. propagating shape over part-of. This could be done on a per attribute basis.

[rays22]

I think I have the clarifications I need.