Update bridging relations in cross-species bridges #3466
Conversation
Update the SSSOM/T-OWL ruleset producing the bridges so that an exact cross-species exact match yields the two following axioms: <subject> EquivalentTo: <object> and (in_taxon some <taxon>) <subject> SubClassOf: part_of some <taxon> (part_of being replaced by occurs_in if the subject is an occurrent) instead of <subject> EquivalentTo: <object> and (part_of some <taxon>) The two-axioms form has always been the intention for those bridges (see discussion in #2428), the one-axiom form that we have used so far was only because of technical limitations.
Now that the equivalence axioms in cross-species bridges use 'in taxon' instead of 'part of' or 'occurs in', the unfolding step when building the composite ontologies need to unfold on that relation as well.
We need 'occurrent part of' as it a parent of FBdv's 'substage of' (FBdv:00018001); without that relation, there is no link between 'substage of' and 'part of', which means that the property chain 'part of' o 'in taxon' SubPropertyOf: 'in taxon' cannot apply to 'substage of' relationships. This is turn means that some relations between FBdv terms cannot be properly translated into class expressions when generating Composite Metazoan.
Update the documentation about the bridges and the composite ontologies to reflect the fact that bridging equivalence axioms now use the 'in taxon' relation, rather than 'part of'.
gouttegd
added
bridge-files
|
This needs a @cmungall glance, I am not a good reviewer for this; I can help find someone that is. |
|
The PR makes complete logical sense, and this was indeed the original
intent.
However, representing this unilaterally in Uberon could cause confusion.
Queries that work for UBERON may fail for GO or CL.
But even assuming we can move forward in unison, I now think there is value
in using different relations for different use cases. Granted this is not
the underlying basis for RO, so there are valid cases either way. Would be
good to gather opinions for others. I can expand on with some examples
later.
One of my concerns is increasing raggedness in the part-of lattice. User
sees 'flexor pollicis longus muscle' as a part of Human in OLS. They open
up Human in OLS expecting to see all human parts. They get confused when
there is just some tiny collection of terms.
…On Thu, Jan 23, 2025 at 3:12 PM Nico Matentzoglu ***@***.***> wrote:
This needs a @cmungall <https://github.com/cmungall> glance, I am not a
good reviewer for this; I can help find someone that is.
—
Reply to this email directly, view it on GitHub
<#3466 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AAAMMOKVE42LB7VC3PFGDQT2MFZMXAVCNFSM6AAAAABVX7NAHWVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDMMJRGIYDEMJQGA>
.
You are receiving this because you were mentioned.Message ID:
***@***.***>
|
Not sure I follow you. The relations updated by this PR would not go into UBERON itself -- they would only go into the bridges (and from there into the collected/composite products). Or are you referring to queries on Ubergraph (which does use the Uberon-provided bridges)? If so, note that this PR also updates the CL bridges - since they are maintained here in Uberon.
I don’t think it is realistic for Uberon to provide several versions of the bridges or of the collected/composite ontologies, each using slightly different axioms to accommodate for potentially different use cases. If people have use cases that require different bridging axioms that the ones we provide in the pre-built bridges, they now have the possibility of downloading the Uberon SSSOM mapping set and generate their own bridges. |
This PR changes the axioms that make up the cross-species bridge ontologies.
Currently, foreign (taxon-specific) terms are bridged to their Uberon counterparts by a single equivalence axiom that looks like this:
or like this (if the term represents an occurrent, such as a developmental stage)
However, this has only been intended as a temporary situation. The original intention was for each foreign term to be bridged by a pair of axioms as follows:
if the term represents a continuant, or
if the term represents an occurrent.
The two-axioms form had been left aside in favour of the single-axiom form solely for technical reasons, because the bridging pipeline at the time was too difficult to update to accommodate for such a change. This is no longer true, so this PR now implements the two-axioms form.
The Composite Metazoan pipeline is updated accordingly to “unfold“ over 'in taxon' rather than over 'part of' or 'occurs in'.
closes #3409
closes #3378