Add bridges between Uberon/CL and taxon-specific anatomy ontologies.

[gouttegd]

This PR adds the Uberon-provided bridges that link Uberon/CL and all the taxon-specific anatomy/development ontologies that are already included in Ubergraph.

The added bridges are for:

• MA
• WBbt/WBls
• ZFA
• MmusDv
• HAO
• EMAPA
• FBdv

closes #136

[balhoff]

@gouttegd in my experience Uberon and EMAPA are not really compatible when logically bridged. Has this been fixed?

[gouttegd]

No. With Uberon’s disjointness axioms left in place, merging Uberon, EMAPA, and the Uberon-to-EMAPA bridge still yields ~1400 unsatisfiable classes.

[gouttegd]

@dosumis @balhoff What is the intention here? Should those bridges be added only after the incompatibilities issues have been fixed?

[balhoff]

Ubergraph is meant to be a logically coherent system, and the relation-graph depends on that. We could add "mapping" versions of the bridges that aren't compatible, such as UBERON:femur skos:closeMatch EMAPA:femur. @dosumis what is your opinion?

[dosumis]

Agree we should keep it outside of the logic if it leads to so many inconsistencies. If we have a compelling use case for including EMAPA, maybe it should use SKOS for the bridge - at least until we have time to work on it. Grant gods willing, we may have funding next year.

[gouttegd]

Note that EMAPA is not the only ontology to be currently incompatible with Uberon:

• Uberon + EMAPA + Uberon-to-EMAPA bridge: 1519 unsats
• Uberon + MA + Uberon-to-MA bridge: 512 unsats
• Uberon + ZFA + Uberon-to-ZFA bridge: 196 unsats
• Uberon + HAO + Uberon-to-HAO bridge: the entire ontology is inconsistent (owl:Thing inferred to be equivalent to owl:Nothing – this seems to be caused by a single bogus mapping, I’ll have a closer look later).

So do we only want the bridges that won’t cause inconsistencies?

[dosumis]

So do we only want the bridges that won’t cause inconsistencies?

I think yes, but would like to work towards fixing for ontologies that are

(a) used in resources we have a compelling case to integrate or where a species specific ontology has clear value in annotation of single cell data (possibly not case for MA, as CL/Uberon are already widely used and should be sufficient (b) the number of unsats suggests the work is manageable.

For now, I think that means ZFA. Also curious about whether we have a mapping from XAO. Potentially useful for community annotating Xenopus scRNAseq data.

I'm happy to have xref mappings for the others - using mappings in Uberon. That could be useful for EMAPA as we may want mappings to this for work at Sanger - it's possible we will have resources for this at some point in the near future if non-trivial.

[gouttegd]

(b) the number of unsats suggests the work is manageable.

For now, I think that means ZFA

Of note, many of unsatisfiability issues with ZFA were spotted a long time ago and seemingly never addressed on either side.

It doesn’t necessarily mean those issues are not manageable, but I’d be wary of using the (relatively low) number of unsats as an indicator of how easy it would be to fix them.

Several of them seem to be the consequence of a fundamental incompatibility between Uberon and ZFA, where what ZFA calls a “head“ corresponds to the head and the neck in Uberon (incompatibility first flagged here, 10 years ago).

Also curious about whether we have a mapping from XAO.

We have ~600 mappings between Uberon and XAO. Using them to bridge the two ontologies yields 208 unsats.

[gouttegd]

Converting to draft as it is going to take a while to fix the unsats between Uberon and the taxon-specific ontologies.