Capturing the upper level for phenotype ontologies

[matentzn]

In this ticket I will summarise the upper level needs for the phenotype ontology space, because we would really like to be sure it to be captured by OBO core. After long deliberations in the phenotype reconciliation effort (representing more than a dozen ontologies and a large chunk of the human and model organism space) we decided to distinguish five levels of granularity in the wider 'anatomical entity' space:

• Whole organism
• Anatomical entity which captures everything from a (native) cell upwards (to what you call gross-anatomical entity)
• Immaterial anatomical entities such as lines (not using them yet, but we will)
• Sub-cellular components (which captures GO:cellular components)
• Molecular entity (CHEBI:23367), for some reason we are still using CHEBI:'Chemical entity' which is more general, but we mean molecular entity

And:

• independent continuant (BFO): this is used as a fall through for capturing everything of the above, to denote a location where something is happening (example).

There are obvious problems especially distinguishing between sub-cellular components and molecular entities, but for now, we are not too fuzzed about clear distinctions. We just need a class that captures 'everything below sub-cellular components', and we dont really mind about what that exactly means.

Other classes we need:

• behavior (currently torn between NBO and GO:behavior process, but we do not care which it will be in the end, as long as all behaviour is captured).
• pathological entity (neoplasm, abscess etc)
• biological process (GO)
• molecular function (GO)
• cell type (CL)
• (phenotypic) quality (PATO)

If you are interested in our patterns, we have a fully (BFO) classified pattern.owl available; all the phenotype patterns can be found under "specifically dependent continuant".

When OBO core is ready we would like to replace all fillers in all patterns by OBO core classes, but expect the same classification as we have it now.

[bpeters42]

Thanks a lot, this is fantastic news. The points you marked as unclear (drawing the line between 'subcellular component' and 'molecular entity'; the need for 'locations') have definitely not been finalized, and it will be very valuable to involve the phenotype folks in these discussions. I expect that especially for the 'line drawing' there is really no right or wrong, but there is a need to come up with something practical.

[matentzn]

So if I want to, say, add behaviour, pathological entity, phenotypic quality, phenotype (not the information item phenotypic finding, who will add them to COB as a result of a ticket, and who decides whether or not to add them?

[cmungall]

I think it may help to distinguish

1. The placement of phenotype
2. The placement of upper level classes on which phenotype ontology definitions depend

1 is hard, but we can at least start a ticket outlining some of the challenges

For 2, you have provided an excellent analysis. I think the correct thing to do here is individual tickets for each of these classes, or to link to existing ongoing discussion. In some cases, COB may work fine (e.g. quality), it might be still good to record this somehow.

One general comment: am surprised at the groups decision regarding excluding subcellular anatomy from anatomy, which is distinct from GO's usage e.g in development, and the phenotype ontologies have been historically well aligned terminologically, as well as terminologically different from CARO which also involves many of the same people (cc @mellybelly @dosumis). Of course we can allow different communities their own terminology but historically this has been a recipe for confusion.

[matentzn]

Ok moved the phenotype ticket here; lets keep this ticket for the remaining upper level classes.

[matentzn]

"excluding subcellular anatomy from anatomy" -> I am not sure we actually do that, or should be doing that if we are.. There are just many patterns that apply only to subcellular components, so we drew this arbitrary line. But since subcellular components can be part of anatomical entities (like cells), that line seems somewhat arbitrary and introducing a big overhead. I think this is also a matter of communication. The patterns of "gross" anatomy and "subcellular" anatomy are actually just compatible - identical in some cases. But some curators where uncomfortable using a pattern that is documented to be used for anatomy in the case of sub-cellular components.

[sbello]

To Chris' point while the patterns for gross anatomy (using an anatomy ontology) and sub-cellular anatomy (using GO terms) are separate, many ontologies place both of these under the morphology branches. The break of morphology vs physiology has mostly been drawn at the molecule/macro-molecule level