Talk:Which relationships should we consider?

From HAO Wiki

Need to consider the implications of each relationship we use. E.g., if A part_of B does that mean that every instance of B has_part A? We should look here for inspiration: http://obo.cvs.sourceforge.net/*checkout*/obo/obo/ontology/OBO_REL/ro.obo

--Ardeans 21:27, 23 January 2009 (UTC)

We should likely just focus on relationships that would be most useful for phylogenetics and taxonomy - at least for now. So, for example, we probably don't need developmentally related relationships (develops from).

--Adeans 14:40, 5 December 2006 (EST)

Would 'is homologous with' replace 'synonym of'?

--Adeans 13:30, 24 October 2006 (EDT)

What about situations where something is 'part of' something but not in all taxa? E.g., the 'apical process of the labrum' is found in Halictidae. Obviously it's a part of the labrum but not in all taxa. Do we need a special relationship ('sometimes part of') for that? Or is simply stating that Halictidae is the highest taxon to which this applies enough info?

--Adeans 17:58, 24 October 2006 (EDT)

I definitely see the utility in adding in relationships like 'adjacent to', especially in aiding image searches in Morphbank. No one wants to include every single body part present in their image. One could simply say 'tarsal claw' if that was the focus of the image. Someone could then search for 'arolium' + adjacent parts and get images tagged simply as 'tarsal claw' but which may also include nice views of the arolium. It's very important that we think about potential relationships now before the ontology gets super huge I think.

--Adeans 16:25, 23 October 2006 (EDT)

The plant ontology uses these relationships:

• develops_from
• disjoint_from
• inverse_of
• is_a
• part_of
• union_of

--Adeans 16:39, 31 December 2007 (EST)

The FlyBase ontology uses these relationships:

• disjoint_from
• inverse_of
• is_a
• union_of

--Adeans 16:46, 31 December 2007 (EST)

More relationships to consider, though some of these perhaps go beyond what we need at this point in time:

• articulates_with <--> articulates_with
• contains <--> is_part_of
• floors
• forms <--> is_formed_by
• has_citation <--> is_cited_by
• invests <--> is invested_by
• is_absent_in
• is_bound_in
• is_dorsal_to <--> is_ventral_to
• is_exposed_in <--> is_not_exposed_in
• is_formed_by <--> forms
• is_fused_to <--> is_fused_to
• is_homologous_with <--> is_homologous_with
• is_in
• is_lateral_to <--> is_medial_to
• is_located_in
• is_not_homologous_with
• is_partially_homologous_with <--> is_partially_homologous_with
• is_present_in
• is_synonym_of
• is_term_preferred_over
• overlies <--> underlies
• overlaps

--Adeans 17:05, 31 December 2007 (EST)

As points of comparison, Flybase and the ontology for mosquito anatomy (two different things!) use six relations (in alphabetical order):

1. develops from
2. disjoint from
3. inverse of
4. is a
5. part of
6. union of

The tick anatomy ontology uses only the last five. Synonyms seem to coded in a different fashion (I'm browsing these using OBO-Edit).

Some comments concerning the relations that Andy's listed in the main article:

• obsolete synonym of: I don't think we should be labelling terms in this manner, we should be amoral on the subject. If the ontology (and resultant glossary) works, then users will be able to determine the meaning of terms that might be unfamiliar.
• is homologous with: How does this differ subtantively from the is_a relationship?
• is larval specific: The three ontologies I've cursorily browsed embed this kind of idea into their classes, i.e., there's a larval midgut and an adult midgut.
• adjacent to: At least with the example provided, adjacency might be inferred by knowing that both scutum and scutellum have the relationship part_of to notum (or thoracic tergum). Adjacent itself, as well as any other words I can think of, is uncomfortably ambiguous.
• not a: This might be the same as disjoint from (I need to figure that out).
• absent in: At least in this example, and following the model from Flybase, wing would be somewhere in the hierarchy of adult anatomy and would not be present in the parallel hierarchy for larval anatomy. Therefore, absent_in seems to be implicit in the relations.
• defining terms: To me, such items would be part of the definition of the term and there's no need for a separate relationship.
• attaches to: The muscle part of Flybase doesn't deal with this at all. For example, coxal tergal remotor muscle 48b, has the chain of is_a relationships: coxal tergal remotor muscle -> indirect flight muscle -> flight muscle -> skeletal muscle of mesothorax -> skeletal muscle of thorax -> adult somatic muscle (followed then by part_of muscle system). Additionally, there is no definition of the term itself. This suggests that they've decided that they can't model absolutely everything, and are doing what's important to the tasks at hand. We should think likewise.
• sometimes form: I don't think that this is an issue. Without thinking about the accuracy of the example, mesofurcal anterior arms may be part_of a mesofurcal bridge (or something like that), but that relationship doesn't imply that all mesofurcal anterior arms form a mesopleural bridge.

--Njohnson 10:07, 4 January 2008 (EST)

I really like and concur with all Norm's of observations. If we accept his vulcan-esque logic, then, just for reference, we are left with the following widely used relationships for consideration (starred entries are presently being used):

1. is a * (larva is a hymenopteran)
2. type of * (areolate is type of surface sculturing)
3. part of * (compound eye is part of head)
4. synonym of * (unguis is a synonym of tarsal claw)
5. develops from
6. disjoint from
7. inverse of
8. union of

And some comments-

• synonym of - Without careful usage this relationship is just as "moral" as "obsolete_syn_of". An example of amoral usage is relating abbreviations to their fully spelled out terms ('not' synonym_of 'notauli'). Most other usage is based on someone's interpretation. This might hint that we need "abbreviation_of"?
• develops from - With the exception of a few cases (egg -> larva etc.) this seems to require information we won't have? Perhaps the best candidate for addition though, as its the simplest way of adding processes to the ontology.
• disjoint from - if Norm's conclusion that this is equivalent to not_a is correct, and that seems logical, then I don't think we need this relationship. It seems that we can just invert the positive representation (is_as) to get negatives. If within the set of relationships within the ontology we can not find that a leg is_a antenna, then we can assume that leg is disjoint_from (not_a) antenna. If, for instance, this example is not true, then we are missing a relationship that should be added.
• inverse of - Can someone provide a meaningful example for morphology? While this might be useful within the Drosophila world where you have knowledge regarding developmental processes I don't see it being useful to us at this time.
• union of - For our purposes I think that this can also be inferred from proper implementation of is_a. Perhaps someone can provide a countering example.

Unless there are other objections I will strike the presently used obsolete_syn_of relationship type from the editor, there are 50 existing relationships that use it, I'll archive them somewhere. I do think that some "social" (not necessarily moral) functionality is needed within the editor/display. We can likely implementing this by individually tagging comments/keywords to terms or relationships.

--MattYoder 14:27, 5 January 2008 (EST)

Perhaps one of the "negative" relationships is for things like "dorsal incision of the propleuron <something> propleuron"? Incisions being the absence of something... --MattYoder 21:50, 7 January 2008 (EST)