Overhaul taxon subsets

[gouttegd]

This PR updates the way we are building “taxon subsets”.

As explained in #3362, we currently have, for reasons unknown (to me at least), two slightly different methods to create taxon subsets: one producing the -view subsets (human-view, mouse-view, xenopus-view) and one producing the -basic subsets (amniote-basic, euarchotonglires-basic). Both methods rely on the use of OWLTools.

This PR replaces both methods by a single one (so that all taxon subsets are produced in the same way) that relies on a new command in Uberon’s custom ROBOT plugin.

The PR does not change which taxon subsets are produced and released by default (the five subsets aforementioned: human, mouse, xenopus, amniote, and euarchotonglires).

More subsets can be produced on demand, all that is needed is to define a TAXON_ID_subsetname Make variable pointing to the desired NCBITaxon ID.

For example, to create a subset for, say, insects, one can do:

sh run.sh make subsets/insect-view.owl TAXON_ID_insect=NCBITaxon:50557

The PR also adds a possibility to create, not a subset directly, but a small component containing only oboInOwl:inSubset annotations to “tag” classes that belong to a taxon subset. For example:

sh run.sh make subsets/human-tags.ofn

would create a human-tags.ofn component containing, for all Uberon classes that belong to the human subset, oboInOwl:inSubset <http://purl.obolibrary.org/obo/uberon/core#human_subset> annotation assertion axioms. Such a component can then be merged with the main ontology for downstream use (e.g., extracting all the classes of the subset).

closes #3362

[matentzn]

I have reviewed the code changes, they look great, - I would like to offer a single word of caution: renaming files, even subset files, may break existing pipelines somewhere on the deep web unless we also add a purl redirect to the OBO purl config.

[matentzn]

Can't imagine what this second relax does but well since it's there..

[gouttegd]

renaming files, even subset files, may break existing pipelines somewhere on the deep web unless we also add a purl redirect to the OBO purl config.

I know. But precisely, this is to be handled at the PURL level, which is here exactly for this purpose. There is no reason for us to refrain from renaming files if it brings better consistency (in this case, by having all taxon subsets consistently named something-view.owl instead of having some subsets being named something-basic).

[dosumis]

Hi @gouttegd - great to see this.

I'm especially excited to see this:

The PR also adds a possibility to create, not a subset directly, but a small component containing only oboInOwl:inSubset annotations to “tag” classes that belong to a taxon subset. For example:

sh run.sh make subsets/human-tags.ofn

would create a human-tags.ofn component containing, for all Uberon classes that belong to the human subset, oboInOwl:inSubset http://purl.obolibrary.org/obo/uberon/core#human_subset annotation assertion axioms. Such a component can then be merged with the main ontology for downstream use (e.g., extracting all the classes of the subset).

I would really like the tags to in incorporated into the release files. We can use them straight away in our autosuggest pipelines to boost for species relevant terms.

[gouttegd]

I would really like the tags to in incorporated into the release files.

I wasn’t sure whether this was your preferred option, so for now the idea was to merely produce the -tags file and leave downstream users of Uberon merge them with uberon.owl if they wanted.

But it certainly can be done directly upstream if preferred.

Do we want all release artefacts to include those tags (e.g. uberon.owl, oberon-basic.owl, but also all the organ-specific subsets such as nephron-minimal, sensory-minimal, etc)? Or just one supplementary artefact containing the taxon subset tags (e.g. something like uberon-with-subset-tags.owl)?

Based on your comment I assume the former (tags included in all release artefacts), in which case we’ll need a new intermediate file (upstream of uberon.owl) from which the subset tags can be generated before we produce the final uberon.owl with the subset tags included.

[gouttegd]

PR updated to include the oboInOwl:inSubset tags directly into the release products.

For now, we include only the tags for the human subset and the mouse subset. To add tags for another subset, it would simply be a matter of adding another tag file to the POSTPROCESS_ADDITIONS Make variable, for example:

POSTPROCESS_ADDITIONS = subsets/human-tags.ofn \
                        subsets/mouse-tags.ofn \
                        subsets/drosophila-tags.ofn

Some considerations:

1. The tags are added as part of the release pipeline, and are not imported from the -edit file. This means the tags will not be visible when working on the ontology from Protégé. This is on purpose. The tags are automatically computed and are not supposed to be manually edited, but there would be no way to enforce that if they were visible from Protégé.

2. The tags are added to the main uberon.owl artefact, and from there are propagated to all files that are derived from this artefact. They are, however, explicitly excluded from Composite-Metazoan (and its progenitor Collected-Metazoan). Including them into Composite-Metazoan would produce a somewhat weird result in that Uberon terms would be tagged as belonging to a given taxon subset, but terms coming from the taxon-specific anatomy ontologies would not be tagged. For example, UBERON’s artery would be tagged as belonging to the mouse subset, but not EMAPA’s superior vesicle artery, even though that term is (obviously, as coming from EMAPA) definitely applicable to mice. I believe this would be confusing.

3. The additions of the mouse and human tags are adding about ~3MB to the main uberon.owl file (from 88MB to 91MB). Not sure whether this is a big concern or not, but I think it should be noted, especially if we want to add more taxon subset tags to the main release file.

[matentzn]

assuming that uberon.owl was swapped out to POST_PROCESS in all the relevant places in the makefile, this looks good to me!

[gouttegd]

assuming that uberon.owl was swapped out to POST_PROCESS in all the relevant places in the makefile

Only two places, in fact:

1. As the input of the Composite-Metazoan pipeline, because as I have said in my last comment, we (at least I) do not want the taxon subset tags to be included in Composite-Metazoan, so it’s fine to build Composite-Metazoan from the POSTPROCESS_SRC intermediate file (which is the same thing as the final uberon.owl minus the taxon subset tags);

2. As the input of the “bridge checks”, because the taxon subset tags are completely unnecessary for those checks (they would cause no harm, but would serve no purpose either).

All other custom pipelines that were dependent on uberon.owl are still dependent on uberon.owl, meaning those custom pipelines now work on a version of the ontology that includes the taxon subset tags. Notably, this means that the “system subsets” (e.g. circulatory-minimal, digestive-minimal, etc.) do contains those tags.

[gouttegd]

Merging here.

@dosumis: (1) If you want more taxon subset tags than just human and mouse to be included in the main product, just say so and we can add new subsets anytime, it’s one line to add in the custom Makefile.

(2) I assume you want to see the same thing in CL?