Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Advice on use of terms from orphaned REX ontology #2002

Open
schaferw opened this issue Jul 28, 2022 · 24 comments
Open

Advice on use of terms from orphaned REX ontology #2002

schaferw opened this issue Jul 28, 2022 · 24 comments
Labels
policy Issues and discussion related to OBO Foundry policies question A question

Comments

@schaferw
Copy link

The REX ontology has a number of terms that are pertinent to PROCO's domain but we are reluctant to just import them as-is since the REX ontology is orphaned. There would be no mechanism for term editing etc. Is it possible to incorporate them into PROCO giving attribution to the original work? We would then assume responsibility for governance of the terms going forward.
@schaferw schaferw added attn: Editorial WG Issues pertinent to editorial activities, such as ontology reviews and principles review request labels Jul 28, 2022
@nataled nataled added policy Issues and discussion related to OBO Foundry policies attn: OFOC call Issue to discuss on fortnightly OBO Operations meeting question A question and removed attn: Editorial WG Issues pertinent to editorial activities, such as ontology reviews and principles review request labels Jul 28, 2022
@nataled nataled removed their assignment Jul 28, 2022
@nataled
Copy link
Contributor

nataled commented Jul 28, 2022

--corrected labels and assignee--

Your question is one that is not fully settled, so far as I know. I'm therefore going to make sure this is discussed at the next OBO Foundry conference call, as it is of wide interest. The call will take place August 9th. If time is more urgent, I suggest soliciting advice from the [email protected] mailing list.

@schaferw
Copy link
Author

schaferw commented Jul 28, 2022 via email

@cmungall
Copy link
Contributor

cmungall commented Aug 9, 2022

Option 1

PROCO (and possibly a consortium of other groups) takes over REX and revives it. This would require the consent of the last owner, Kirill, and this should be done openly on obo-discuss to give any current users of REX the option to speak up with any concerns, but I think it should be fine. There would be an initial cost involved in setting up a modern ODK repo for REX that would make it easy for others to contribute.

Option 2

A new ontology or new branch of PROCO is created for physico-chemical processes. REX terms are copied there as needed but with a new ID mintred, with full provenance pointing back to the source.

What you should not do is reuse the REX ID and modify/inject axioms

General concerns

I am unclear on the scope differences between MOP and PROCO. Additionally I see PROCO is importing terms from allotrope, and including very general terms like "state" and placing them under BFO process. I think for general terms like state these should go via COB.

@pbuttigieg
Copy link
Contributor

What's the status of PROCO? This branch looks concerning, with functions and states under processes.

@balhoff
Copy link
Contributor

balhoff commented Aug 9, 2022

What's the status of PROCO? This branch looks concerning, with functions and states under processes.

If you mean GO 'molecular_function', it is supposed to be under 'process'.

@nataled
Copy link
Contributor

nataled commented Aug 10, 2022

@schaferw here is the outcome of the meeting and the current status. During the call the options presented above were discussed. The general consensus is (1) Option 2 is more lightweight (and was the suggested/preferred option); (2) the option chosen will largely depend on how much of REX is needed (and thus the choice is yours as to which route to take); and (3) in either case an attempt to reach out to the REX developers should be made. A member of our team that is in contact with said developers has already done so on your behalf. We await a response.

@schaferw
Copy link
Author

schaferw commented Aug 10, 2022 via email

@matentzn
Copy link
Contributor

I think limited import is really a problem without deprecting REX entirely. And if deprecate the whole ontology, we will lose a lot of potentially interesting terms. My preference would be to get the REX people to add you as editors, open REX up for community edits and you then importing whatever you need into PROCO straight from REX. Deprecting an ontology without replacement feels wrong as well.

@nataled
Copy link
Contributor

nataled commented Aug 10, 2022

@matentzn many ontologies have been deprecated without replacement, that's just the way of it sometimes. Terms that are needed from a deprecated ontology are typically rescued, I suppose, by those that (a) know how to maintain those terms, and (b) have a need for them. Those not rescued are left untouched for a reason--not needed, or perhaps not well formed.

As for limited import without deprecating, that is potentially a problem (hence my concern on the call about the possibility of an orphaned ontology becoming reactivated). The solution we are pursuing is exactly to deprecate, but to first ask about doing so.

@matentzn
Copy link
Contributor

Yeah I reached out to the maintainers and will fill you in what they say. Actually FYI: it seems the Chebi team is responsible for REX now.

@K-r-ll
Copy link

K-r-ll commented Aug 10, 2022

Hello people,

What a surprise! I was approached by the ChEBI team who asked me what I wanted them to do about REX. Somehow I missed the fact that they were even responsible for maintaining it. Apparently the ChEBI team inherited the contact after the EBI deleted my email account. So... I suppose the aforementioned (in this thread) "REX developers" must be me.

Anyway. At the time it was created in the early 2000s, there was little, if any, enthusiasm at the EBI, and not much after that either. There was some vague and non-committal interest from the Royal Society of Chemistry because at some point they used to cross-reference their publications with REX terms. The last changes to REX were made (by me) more than a decade ago which means it's about time to move on.

Sure it would be nice if anybody gave it a new life, maybe (just maybe) preserving some of its original ideas form whatever. For me, @cmungall's option 1 sounds more attractive. I also don't think it is incompatible with Option 2. Nothing should prevent the users to keep recycling the REX terms as they please.

If you have any questions about those terms themselves, feel free to ask; not sure if by now I can answer though.

@schaferw schaferw mentioned this issue Aug 11, 2022
Open
@schaferw
Copy link
Author

schaferw commented Aug 11, 2022 via email

@schaferw
Copy link
Author

schaferw commented Aug 11, 2022 via email

@schaferw
Copy link
Author

schaferw commented Aug 11, 2022 via email

@K-r-ll
Copy link

K-r-ll commented Aug 11, 2022

Further to my above post. Mr Google brought me a thread from 2014 which sheds some light on interaction of CHMO, REX and FIX: https://code.google.com/archive/p/obo-foundry-operations-committee/issues/128

Janna Hastings said:

We do want to keep these alive at least until we have replacement ontologies in the same domain. <...> I've just moved the sources to the ChEBI webspace as these were originally in Kirill's home drive in EBI which has since been removed as he left EBI some years back.

Curiously, or maybe not, on the top of the same thread, Colin Batchelor said:

So I think the story is to keep REX and I'll make sure that everything in FIX has been put in CHMO with appropriate connections to OBI and then we're fine.

So it seems that the RSC people at least intended to recycle everything of FIX. However it does not look like they kept the corresponding FIX IDs, as claimed:

There are quite a few cross-references in CHMO to FIX; Hilary was diligent about this but she may have registered them in a slightly eccentric way.

For instance, cf. momentum-resolved Bremsstrahlung spectroscopy FIX:0000688 and momentum-resolved bremsstrahlung spectroscopy CHMO:0000380

@cmungall
Copy link
Contributor

cmungall commented Aug 12, 2022 via email

@pbuttigieg pbuttigieg mentioned this issue Aug 12, 2022
Open
@matentzn
Copy link
Contributor

@cmungall can you summarise what your opinion is on the question of whether it is acceptable to adopt terms from a non-obsolete ontology? It seems odd to me. I think REX should either be maintained as a whole, or obsoleted, but allowing other ontology groups to grabbing REX terms, axiom inject, change labels etc etc, while REX is still active, seems against OBO principles

@cmungall
Copy link
Contributor

cmungall commented Aug 24, 2022 via email

@pbuttigieg
Copy link
Contributor

An observation:

Many rounds of rescue will (and has) split generic/broadly applicable content across other ontologies with narrow(er) scope. This also happens when an ontology with a narrow(er) focus decides to handle very generic content (e.g. plans), which should be maintained as its own (relatively static) ontology.

This makes the overhead importing the natural set of terms (now scattered across ontologies) a headache or intractable to import and coordinate, and also subject to idiosyncracies from the narrower foci.

"Pragmatism"

If it were easier to edit multiple ontologies and sync editor versions (syncing release cycles is non-viable at this stage), it would be much easier to co-maintain generic ontologies. That is the change that will improve (inter)operations sustainably.

Right now, I would just create terms in ENVO (even those outside its scope) rather than dealing with handling import and integration of scattered terms that should have their own generic ontology.

@StroemPhi
Copy link

StroemPhi commented Aug 29, 2022 •
edited
Loading

When comparing REX and MOP, one can see a great overlap between the two (e.g. catalysis or oxidation) that unfortunately is not marked with annotations in MOP. In terms of scope and orthogonality this is a problem. I also found the 2014 thread Kiril referenced in his #2002 (comment) and thus assumed, since REX and FIX are orphaned on the OBO page, that MOP, RXNO and CHMO are now the place to look for terms in this scope/domain. It would definitely help to have a better comment/explanaition on what 'orphaned' means. I interpreted it as: it's listed here for backwards compability reasons but you should better not use it in new projects, as it is not maintained anymore and instead use maintained ontologies with similar scope/domain.

Now, if REX were to be reactivated the overlap with MOP would need to be resolved, otherwise users will get really frustrated by not knowing which one to take and there would be a broken OBO rule.

To me, it seems that MOP would be the better place to have those terms of REX that don't have a MOP equivalent already, as @batchelorc is actively maintaing MOP, REX has many missing definitions and has less BFO structure than MOP. But importing the tems from REX, if reactivated, via ODK might also work. For the latter though, I think REX would have to be vetted regarding its subsumption hierarchy (e.g. 'oxidation' and 'reduction' being siblings of 'process', which is not bfo'process').

IMHO, I think it would be best to combine the efforts to harmonize REX and MOP to have a proper reference ontology dealing with the most basic chemical processes. A couple of month ago, Colin and I untangled MOP and RXNO with the goal to have both seperatly in an ODK repo and then properly import MOP into RXNO. I came quite far already but then had to switch to tasks with a higher priority unfortunately. So having MOP in a new ODK repo, might probably be easier to achieve than for REX.

If there where an easy to use scipt (@cthoyt?) with which we could have similar terms - judged by label, definitions incl def source links (often IUPAC Goldbook) - of REX on the side and MOP on the other of a table, I think the needed harmonization could be done manually with the help of domain experts/chemists. With this, we could decide to deprecate those REX terms that have a proper MOP equivalent and just import or adapt the others (if the hierarchy allows it) in/to MOP, or at least make a SSSOM mapping between the two.

(Am really sad, that I stumbled across this issue thread just now!)

@cthoyt
Copy link
Collaborator

cthoyt commented Aug 29, 2022

Yup I’ve made scripts like this before. Even more, I’m familiar with the overlaps between these ontologies, sort of related to Chris’s recent blog post about shadow ontologies

@cmungall
Copy link
Contributor

I like the path emerging here, a unified chemical process ontology seems like the way to do

@cthoyt
Copy link
Collaborator

cthoyt commented Aug 29, 2022

I did a lot of curation in that direction to support pybel in the past, would be very interested to be involved in or lead that effort, assuming we could do all curation in TSVs :)

@nlharris
Copy link
Contributor

nlharris commented Sep 4, 2022

Is there more to be discussed here, or has the original question been answered sufficiently?

@nlharris nlharris removed the attn: OFOC call Issue to discuss on fortnightly OBO Operations meeting label Sep 6, 2022
@pbuttigieg pbuttigieg mentioned this issue Sep 6, 2022
Closed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
policy Issues and discussion related to OBO Foundry policies question A question
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
10 participants
@cmungall @balhoff @nlharris @pbuttigieg @cthoyt @matentzn @nataled @StroemPhi @K-r-ll @schaferw