oborel / obo-relations

RO is an ontology of relations for use with biological ontologies
http://oborel.github.io/
Other
92 stars 47 forks source link

NTR: [sourced,extracted,...] from or variation of derives from #621

Open pbuttigieg opened 2 years ago

pbuttigieg commented 2 years ago

RO:'derives from' requires that the old entity ceases to exist. What relation should we use for things that have been extracted from an entity that continues to exist?

e.g. a biopsy of liver tissue, a bucket of lake water, a portion of soil from a field, a grain of sand taken away from a beach, etc

pbuttigieg commented 2 years ago

I'm creating "derives from part of" in ENVO for the moment.

I only see "derives from" in RO, and then "derived by..." classes

jamesaoverton commented 2 years ago

I was working on something similar for another project, and thought this note might help someone down the line.

OBI has various 'specimen' classes that we want to link to their sources, and we've used 'derives from': e.g. 'blood specimen' 'derives from' from some UBERON 'blood' from some particular human. I think that the composition 'derives from' * 'part of' works in this case as long as we (correctly) understand that it isn't all the blood in the person, but just a small part. So this RDF (Turtle with labels) should work:

:specimen-1
  a 'blood specimen' ;
  'derives from' [
    a 'blood' ;
    'part of' :james
  ] .

I think this works for blood and other bodily fluids, but probably not for our 'liver specimen', since it implies that the liver ceased to exist, as @pbuttigieg is pointing out. We could talk about 'liver tissue' instead, but UBERON doesn't (seem to) have that class. Or we could add a second 'part of' step:

:specimen-1
  a 'liver specimen' ;
  'derives from' [
    a 'tissue' ;
    'part of' [
      a 'liver' ;
      'part of' :james
    ]
  ] .

Having one step for "mass nouns" and two steps for "count nouns" makes modelling and querying harder.

I can see the value of 'derives from part of' relation, and maybe this helps clarify it slightly.

ddooley commented 1 year ago

In line with Pier's request, could we please have an "separated from" object property between material entities! In the case of anatomical or substance part-hood, it would allow us to know that the separated animal or material entity part has the same composition as described by anatomy or say geological reference, but sans contextual connections.

label: separated from definition: A relation between a material entity which has been separated from another material entity.. example of use: A liver tissue biopsy was separated from a liver. A leg was separated from an animal. An apple was separated from a tree. parent: [top level?]

"Separated" seems ambiguous enough about connectivity to catch the cases we want. But we could use removed perhaps. Terms such as severed, disconnected or extracted hint more specifically at kinds of processing involved. Terms such as "amputated", "eviscerated", "harvested" would specialize the kinds of things removed.

Use cases:

I understand the "derives from" ceases-to-exist semantic regarding the source (domain) in terms of pure identity but we want something that sidesteps the pure identity requirement, but maintains the composition of the original in-situ part.

ddooley commented 1 year ago

Slightly chagrined - RO currently has "has substance removed", a relation added years ago. Current defn: A relation existing between two physical entities in which the first entity has had the second entity (a substance) removed from it at some point in time. "separated from" is pretty much synonymous with the inverse of this relation!

So can we add "separated from":

label: "separated from"
definition A relation between a material entity which has been separated from another material entity.
inverse: "separated into"

And then tweak "has substance removed" relation:

rename "has substance removed" -> "**separated into**"
tweak definition:  Inverse of "separated from"
inverse: "separated from"

This will line up nicely with OBI's existing "material component separation" processing term!

ddooley commented 1 year ago

We could try "[ME1] 'extracted part' some [ME2]" to convey that ME1 is the entirety of ME2, but removed from ME2's in-situ context. So "'extracted heart' 'extracted part' some [UBERON] heart". But there may be a problem with referring to its parts, https://github.com/OBOFoundry/COB/issues/204#issuecomment-1491243946

wdduncan commented 1 year ago

Not sure if this relevant, but in the OHD we have the term extracted tooth (http://purl.obolibrary.org/obo/OHD_0000233). For some reason, ontobee isn't picking this class up, but here is a screenshot of the class and its axioms.

image