[WIP] OBO format version 1.4 compliance

[althonos]

Hi !

I am currently working with @cmungall on improving the syntax and semantics of the OBO ontology language. This PR is an effort to standardize the syntax in envo-edit.owl so that the produced OBO is compliant with the specification.

The syntax compliance was checked with althonos/fastobo which is an in-development parser, serializer and validator for the OBO 1.4 format.

This is still work in progress, but I opened this issue as a place to discuss some of the requested modifications.

Required

• [x] Fix whitespaces in PMID database xrefs.
• [x] Remove placeholder and/or ill-formatted database xrefs
• [x] Replace PMID for application instance as it is an invalid identifier prefix. This one needs discussion: should it be just replaced with PMID, losing the idea of application instance ? Or should it be moved to a custom property_value ?

Suggested

• [ ] Use the replaced_by clause to indicate the accurate MOD mapping for obsolete terms. If no exact mapping is available, then the consider clause could be used.
• [ ] Use property_value clauses for id-validation-regexp: and search-url: as they are not real xrefs.

[pporrasebi]

Whitespaces and '' issues fixed, I will do some consultation for the 'PMID for application instance' issue and the suggested issues.

[althonos]

Great! Don't hesistate to @ me in relevant issues if any.

[pporrasebi]

Comments on the pending issues for @althonos to check:

• PMID for application instance will be replaced with PMID only.
• replaced_by pointing to MOD in obsolete terms: We are fine with doing this, but unsure as to where exactly you need us to do it. We are still using OBOedit to manage the ontology and it is not clear for us how to implement it.
• search-url: and id-validation-regexp:: For the time being at least we need to leave them as cross-references, since several pipelines expect them to be in that specific field. As in the previous case, it is not clear for us how to implement the property_value annotation uisng OBOedit. Some assitance here would be really welcomed.

[althonos]

(cc @cmungall)

replaced_by pointing to MOD in obsolete terms: We are fine with doing this, but unsure as to where exactly you need us to do it. We are still using OBOedit to manage the ontology and it is not clear for us how to implement it.

The OBO format provides two clauses to help remapping obsolete entities:

• replaced_by, which can be used to automatically remap entities referring to the obsolete entity with the new one. For instance, MI:0154 has been replaced by MOD:00115, and any ontology can safely replace any occurrence of MI:0154 to MOD:00115. Adding a replaced_by: MOD:00115 clause to MI:0154 would allow most tools to do that automatically.
• consider, which can be used to indicate a possible remap that cannot be applied automatically. For instance, MI:0021 is obsolete and MI:0428 should be used instead, but in conjunction with other terms, hence requiring some curation.

With OBOedit however I'm almost sure this is not possible to do, but this can be done in Protege, or with a manual edit in the OBO file.

search-url: and id-validation-regexp:: For the time being at least we need to leave them as cross-references, since several pipelines expect them to be in that specific field. As in the previous case, it is not clear for us how to implement the property_value annotation uisng OBOedit. Some assitance here would be really welcomed.

Once again this is probably not feasible with OBOedit. In Protege this would also be done by defining an AnnotationProperty for each annotation you're currently using xrefs for (e.g. has_search_url) and then adding annotations to the relevant terms.

EDIT:

PMID for application instance will be replaced with PMID only.

The other option is to use a property_value (again) using IAO:0000112 as the relationship and PMID:..... as the annotation value.

[pporrasebi]

After some discussion, we have replaced PMID for application instance with simply PMID, so all required fixes are done.

Regarding the suggested issues, there are downstream pipelines that depend on finding search-url: and id-validation-regexp: as xrefs and changing them involves quite some work on our part and also by third parties. It would be needed to fully understand the problem that having these as xrefs creates before committing to put significant effort adapting existing tools.

Apart from that, we have a problem with using Protege instead of OBOedit and it is its inability to display 'is part' relationships, which abound in our ontology. Navigating the ontology using Protege is not very practical, we would refrain to start doing our developing there until this issue is solved.