Restructuring ‘inflammation’ terms in EFO

[paolaroncaglia]

@cmungall @nicolevasilevsky @pnrobinson Please let us know if you have any comments on the high-level changes in EFO below.

Background: EFO has multiple terms related to inflammation. We aim at avoiding duplicates and improving representation of inflammation-related terms in EFO.

Details of current high-level inflammation-related terms in EFO:

1) inflammatory response (http://purl.obolibrary.org/obo/GO_0006954) This is a GO term (child of ‘response to stimulus’) and is in the EFO process branch. It has the gwas_trait true annotation, meaning it’s being used in GWAS, so we’d like to retain it. It has 2 children (1 with further descendants). These are general terms, not related to specific diseases/disorders.

2) inflammation (http://purl.obolibrary.org/obo/MP_0001845) This is an MP term (child of ‘abnormality of the immune system’) and is in the phenotype branch of EFO. In the original ontology, this term is actually named ‘abnormal inflammatory response’. It has 6 children (some with further descendants). These are all related to specific diseases/disorders.

3) inflammation (MPATH) (http://purl.obolibrary.org/obo/MPATH_212) This is an MPATH (mouse pathology) term (child of ‘pathological entity’) and is in the material entity branch of EFO. It has no children.

We propose to do the following high-level changes:

• [ ] Add EFO:NEW ‘inflammatory disease’ Subclass of disease “A disease involving an inflammatory response in an anatomical structure.” disease_has_feature EFO GO ‘inflammatory response’ OR has_part EFO GO ‘inflammatory response’ definition_citation: MONDO:0021166 Synonym: inflammatory disorder

• [ ] Retain the current term EFO GO ‘inflammatory response’ and its children, but do not place or add disease-specific terms under it; add an editor note to that extent.

• [ ] Rehouse children of EFO MP ‘inflammation’ under the disease branch, according to both their anatomical location and as children/descendants of EFO:NEW ‘inflammatory disease’.

• [ ] Then obsolete EFO MP ‘inflammation’. Do not replace it with another term, but, in the obsoletion note (reason_for_obsolescence in EFO), suggest considering EFO GO ‘inflammatory response’ or EFO:NEW inflammatory disease or one of their descendants.

• [ ] Obsolete EFO MPATH ‘inflammation (MPATH)’ in EFO. Do not replace it with another term, but, in the obsoletion note (reason_for_obsolescence in EFO), suggest considering EFO GO ‘inflammatory response’ or EFO:NEW inflammatory disease or one of their descendants.

(A search for the exact string ‘inflammation’ in EFO would then return only EFO GO ‘inflammatory response’.)

Other action items, of EFO interest only:

• [ ] Inform GWAS of the changes, for their information. In fact, may be worth mentioning changes in EFO release notes + email.

• [x] Add EFO:NEW ‘inflammatory disease’ to MONDO mapping files

• [ ] If we do add the axiom EFO:NEW ‘inflammatory disease’ disease_has_feature EFO GO ‘inflammatory response’, consider editing the definition of the object property disease_has_feature, from the current "A relationship between a disease and some feature of that disease, where the feature is either a phenotype or an isolated disease." to the slightly broader "A relationship between a disease and some feature of that disease, where the feature is either a process, a phenotype or an isolated disease."

• [ ] Enrich and improve the branch under EFO:NEW ‘inflammatory disease’ by adding it as a parent of disease terms that already exist in EFO and contain ‘inflammation’ or ‘inflammatory’ in its definition or label, e.g. 'nephritis' and 'labyrinthitis'. This may be simpler in EFO 3, as we expect that the addition of EFO:NEW ‘inflammatory disease’ mapped to the equivalent MONDO term, and the following integration with the MONDO hierarchy that already has many subclasses of inflammatory disease, will add ‘inflammatory disease’ as a parent where applicable and appropriate.

[paolaroncaglia]

Note for self: David OS and Nico will take a look (probably week after next due to travelling) and I'll follow up with them on or around March 7th. [I emailed them on March 7th.]

[nicolevasilevsky]

@paolaroncaglia I haven't had a chance to look at this either, but I'll aim to do it next week too.

[paolaroncaglia]

@nicolevasilevsky thanks, looking forward to your comments!

[nicolevasilevsky]

@paolaroncaglia this proposal seems fine to me

[paolaroncaglia]

@nicolevasilevsky thanks for your feedback.

[dosumis]

1. I think we need to discuss the relationship between the MP term and the disease term. This should, ideally, align with broader standards.
2. The MP formal def will likely link to the GO term via inheres_in - could EFO simply import this to link to GO?

I'll discuss both of these with @matentzn today.

[dosumis]

Some suggestions following discussion (with the caveat that we probably don't understand EFO and its usage well enough).

1. We agree that 'inflammatory disease' should be added and linked to the MONDO equivalent (in whatever way is currently standard for EFO). There are a bunch of other inflammatory disease terms in EFO that should be classified under this term (e.g. inflammatory skin disease). This could potentially be done with a pattern + inference.

• Some other examples:

2. Presumably all the specific instances of inflammation you want to be able to refer to with EFO are symptoms/phenotypes rather than the normal GO process. We think that means it's better to continue to use phenotype terms rather than GO. If you follow our advice on this, we think you should use the inflammatory disease: 'disease_has_feature' relationship to link disease terms to phenotype terms - starting with MP:'abnormal inflammatory response'. In the glorious EFO3 future this should would ideally be a uPheno term (it should be easy to co-ordinate these with Nico - or just add them directly by editing a uPheno term csv). For now it seems reasonable to keep HP/MP terms for this purpose as they can easily be mapped up to uPheno terms once they are available. The MP term has a formal definition that links to GO.

https://www.ebi.ac.uk/ols/ontologies/mp/terms?iri=http%3A%2F%2Fpurl.obolibrary.org%2Fobo%2FMP_0001845

In EFO3 presumably this could be brought in automatically, but for now this can be added done manually. There seems to be precedent for this in EFO, e.g. this comes from HP.

Caveat: Are all imported terms available to curators? Will having GO and MP therefore cause problems?

3. Some of the subclasses under inflammation have truly terrifying formalisation that need to be fixed.

This means: (i) Every (process) that occurs in an eye is a subclass of 'eye inflammation' (ii). Every (process) that occurs in an eye is a subclass of 'inflammation'

We think that you should try to use MP or HP terms for these, or make EFO terms that use the same definition patterns as MP/HP terms, referencing the relevant. (If you don't follow out suggestion of using phenotype terms, these should at least be fixed to:

eye inflammation EquivalentTo: GO:inflammation and occurs_in some eye

A general point re:

Then obsolete EFO MP ‘inflammation’. Do not replace it with another term, but, in the obsoletion note (reason_for_obsolescence in EFO), suggest considering EFO GO ‘inflammatory response’ or EFO:NEW inflammatory disease or one of their descendants.

Obsolete EFO MPATH ‘inflammation (MPATH)’ in EFO. Do not replace it with another term, but, in the obsoletion note (reason_for_obsolescence in EFO), suggest considering EFO GO ‘inflammatory response’ or EFO:NEW inflammatory disease or one of their descendants.

It's not a good idea to obsolete anything with a foreign ID in EFO. If this assertion gets out into the wider ecosystem of cross-importing ontologies it will cause problems. It would be better to simply delete, or If that doesn't work for EFO workflows, to find some other way to tag as no longer valid in EFO.

[paolaroncaglia]

@dosumis @matentzn thanks for your precious feedback (and @cmungall thanks for your approval). I will re-write the proposal taking into account your suggestions, and will address concerns (together with other EFO team members). For now, a few quick notes:

• [ ] If we retain the MP term, we should update it to reflect its current annotation properties + axioms as they are in MP exactly;
• [ ] Re. danger of obsoleting non-EFO terms in EFO: the MP term would probably stay; the MPATH term could stay but with an added annotation property of editor_note to say that it is no longer valid and indicating what inflammation term should be used instead.

[matentzn]

Note to self: if MPATH terms should in general not be used by curators, we should probably use the new 'namespaces of interest' filter feature in ROBOT to get rid of all of them for the release.

[paolaroncaglia]

@dosumis @matentzn FYI, we will address point 3 from your comment https://github.com/EBISPOT/efo/issues/372#issuecomment-470908980 as part of a separate ticket, see https://github.com/EBISPOT/efo/issues/295#issuecomment-477124283.

[paolaroncaglia]

Note for self: re. MONDO representation of 'inflammatory disease': see https://github.com/monarch-initiative/mondo/issues/653

[paolaroncaglia]

Note for self: Coming back to this and addressing high-level classes first: MONDO:0021166 'inflammatory disease' is now in EFO3; object property 'disease has inflammation site' is now in EFO (https://github.com/EBISPOT/efo/issues/406). So, start by

• [x] adding MONDO:0021166 'inflammatory disease' in EFO2 as subclass of disease "A disease involving a pathogenic inflammatory response in an anatomical structure. [ MONDO:patterns/inflammatory_disease_by_site ] MONDO:0021166 equivalentTo: disease and (has_inflammation_site some ('anatomical structure' or (part_of some 'anatomical structure'))) EFO:0009903 inflammatory disease BUT
• [x] Looks like the nearest term in EFO for UBERON 'anatomical structure' (http://purl.obolibrary.org/obo/UBERON_0000061) is EFO:0000787 'animal component'. Check to avoid downstream complications: https://github.com/EBISPOT/efo/issues/439
• [x] Add to mapping files
• [x] Note that the EFO definition for 'inflammatory disease' is an improvement on the MONDO one ("An inflammatory disease involving a pathogenic inflammatory response in the anatomical structure. [ MONDO:patterns/inflammatory_disease_by_site ]"). I'll open a MONDO ticket to suggest improving their def. But for now, both definitions will live in EFO, and that won't break anything.

[paolaroncaglia]

This is now a duplicate of #471.