search

obophenotype / uberon

An ontology of gross anatomy covering metazoa. Works in concert with https://github.com/obophenotype/cell-ontology
http://obophenotype.github.io/uberon/
Other
132 stars 29 forks source link

'material anatomical entity' subClassOf homologous_in some Vertebrata #246

Closed hlapp closed 11 years ago

hlapp commented 11 years ago

Protege reports this as asserted in merged.owl, and as coming from dc:source VHOG:0001721. Surely this is way too broad? (Though with homologous_in being an undefined property, it could of course mean anything. Still, this will be rather confusing?)

hlapp commented 11 years ago

It actually also leads to some funny assertions, such as haltere only_in_taxon some Arthropoda and homologous_in some Vertebrata. Clearly, this nonsense can be tested for automatically?

cmungall commented 11 years ago

A few things:

cmungall commented 11 years ago

Check VHOGs without supporting evidence

The following VHOGs were used to make homologous_in statements without supporting evidence in organ_association_cHOG table - verify that these are homologous groups. Most seem reasonable but vessels, esp. carotid needs checked.

mellybelly commented 11 years ago

To discuss- consider having the homology assertions in a separate file that imports Uberon. Also To Do: evaluate evidence codes

hlapp commented 11 years ago

I agree with @mellybelly . Many of these classes will have subclasses. For example, do we really want to assume that all kinds of veins will be homologous across vertebrates? Of course, making it an annotation property will prevent propagation to subclasses. But if there is no supporting evidence, what is the point of such an annotation property? How would it be used?

cmungall commented 11 years ago

We're mixing multiple things here:

Modularity and inferences aren't completely separable. For example, the definition of structures like manual digit 3 is tied in with homology.

Inferences: we wouldn't infer all veins are homologous to all veins, this would be incorrect.

Supporting evidence: in vHOG this is annotated as "well established", and a supporting citation: "The appearance of Chordata and subsequently the vertebrates is accompanied by a rapid structural diversification of this primitive linear heart: looping, unidirectional circulation, an enclosed vasculature, and the conduction system." [DOI:10.1196/annals.1341.002 "Bishopric NH, Evolution of the heart from bacteria to man. Annals of the New York Academy of Sciences (2006)"

Utility:

Probably best next discussed on a homology call

cmungall commented 11 years ago

Hmm - there were 3 false positives in my over-sensitive report above. In fact vein, tooth and rectum are in the vhog associations file. You just happened to pick one of these, hence the confusion over evidence.

cmungall commented 11 years ago

related to evidence types, vhog has a confidence level which is useful. this could be formalized

hlapp commented 11 years ago

@cmungall Indeed. Do you want to suggest this to Paula et al re: applying it to the curated set of homology assertions?

fbastian commented 11 years ago

Hi,

FYI, we are currently moving away from vHOG, and transferring our homology statements to Uberon in an external file (very similar to GOAF format). Some of the problems you spot appeared because some vHOG terms were created only to structure the ontology, without any real biological meaning. Those will not be transfered.

We are now working to provide for each homology assertion proper evidence codes (mostly the ones added for phenoscape), and also confidence codes (http://wiki.isb-sib.ch/biocuration/Confidence_information_draft). The homology statements will be provided with taxon information (i.e., the putative ancestor where the structure appeared).

Also, regarding this point: "do we really want to assume that all kinds of veins will be homologous across vertebrates?" I agree that homology assertions should not be propagated to subclasses.

So, as a conclusion:

I take the opportunity to ask: we'd like to request a new evidence code, for homology assertions based on the phylogenetic distribution of a feature (because for now, the term ECO:0000080 only have child terms related to genes). Any thought about that?

hlapp commented 11 years ago

@cmungall where are we this? and are the last two commits (909cf77 and 909cf77) still related to this? (I couldn't quite figure out how Github made the connection.)

cmungall commented 11 years ago

Hilmar -regarding the seemingly spurious commit messages, see #313

Here is where we are:

hlapp commented 11 years ago

@cmungall I see. So does this look ready to close, or what actions are left to be done, discussed, or decided?