Request for 'capable of' predicate for anatomy - GO MF/BP associations about biolink-model HOT 11 CLOSED

The developing min predicate set proposed in #55 includes an ‘enables’ predicate, which was created to support GO annotations of genes that link them to molecular functions they are capable of. A ‘capable of’ predicate was considered here, but for reasons I won’t go into GO chose to use the RO:’enables’ relation in their data instead of its more general parent RO:‘capable of’ relation (the curator note here explains rationale for this).

For Translator purposes, I propose a ‘capable of’ predicate for the Translator minimal set, instead of the more constrained ‘enables’ predicate. KGs can use this ‘capable of’ predicate for connecting things like genes or anatomical entities to GO molecular functions or biological processes.

Thoughts? (in particular from @cmungall)

from biolink-model.

Matt, we are fine with "enables" for the time being. But we are also fine with "capable_of". Basically we just want to implement the standard. So it sounds like we can safely use "enables' for the time being, until such time as there is agreement from the standards committee to allow "capable_of"?

from biolink-model.

Would be great if "capable_of" was at least in the maximal predicate set (it may already be, I confess I haven't checked)

from biolink-model.

from biolink-model.

@sramsey I think enables is too strict for anatomy - GO:BP/MF associations - or at least sounds funny to me. So I'm in favor of generalizing it a bit to 'capable of' or 'involved in', and having this be a predicate in the minimal set.

@cmungall It makes sense to me to have 'capable of' and/or 'involved in' live under the existing 'participates in' blm predicate. Currently we map ro:'involved' in to blm:'participates in'. Are you proposing adding an 'involved in' predicate as another slot under 'participates in' - where 'involved in' would also imply agency but for only part of the process or function?

A straw man for a possible hierarchy of predicates is below (note that it assumes we can conflate BP and MF as possible ranges for all of these predicates - which i think has historical precedent on other efforts):

'participates in'                (any type of participation in a process or function)
      'actively involved in'     (has agency for at least part of the process or function)     
            'is capable of'         (has agency on its own to drive entire process or function)

This seems consistent with the spirit of what Chris proposed, and with what is in the RO. Although in the RO things are a bit more complicated/confounded by ontological considerations (and side note - looking at RO and it is not clear to me how 'involved in' is different from the 'capable of part of' relation.)

If we agree on this simple hierarchy, what should be part of the minimal spec? I think 'participates in' for sure. And i see value in at least one predicate to indicate an agency role in this participation. Not sure we need to distinguish between partial and autonomous agency (i.e. 'involved in' and 'capable of'). This finer grained distinction could be specified if desired using RO relations for the max predicate edge property.

from biolink-model.

To keep things moving ahead, I am going ahead and implementing as proposed above, specifically:

'participates in'                (any type of participation in a process or function)
      'actively involved in'     (has agency for at least part of the process or function)     
            'is capable of'         (has agency on its own to drive entire process or function)

Happy to revisit and revise if this ends up being problematic - but want to get certain predicates in the spec that teams need now for their KGs.

from biolink-model.

One important factor to consider: a gene product M that has an activity that acts upstream of a process P (either regulating P or playing some role that causally affecting P) is not considered a participant in that process in RO. The RO property acts upstream of or within covers the more general case

from biolink-model.

Good point Chris. There are many nuances to consider here, as evidenced by the number and fine-grained distinctions made between the many RO relations in this area.

If we keep them, should be clear that the relations above should be used when the entity does indeed participate in the indicated process or function. If there are cases where we cannot or don't want to distinguish whether the entity facilitates a process or function through upstream action or through direct participation/execution, we can add an 'acts upstream of or within' predicate. But I would do so only if/when there is a real need/use case for it in our work.

For the time being, are there concerns with the three predicates proposed above?

from biolink-model.

copy that. we are implementing is_capable_of predicates to relate proteins to "molecular function" ontology terms.

from biolink-model.

Can this be closed?

from biolink-model.

Yes. The request has been implemented in the model.

from biolink-model.