Infer involved_in Reg of BP_X from MF -R-> BP_X

[dosumis]

In GO-CAM, we commonly have :

• GP enables MF1 -directly_regulates -> BP1
• e.g. FOXO3 enables TF activity, sequence specific DNA binding directly_activates transcription

Based on these, ideally we would infer:

• GP involved_in Reg of BP1
• FOXO3 involved_in regulation of transcription...

OR we would have equivalent MF-BP part_of links in the ontology (involved_in decomposes to enables o part_of)

Solution1 : Intermediate MF term

'transcription regulator activity'
EquivalentTo: molecular_function that regulates some transcription
SubClassOf: part_of some regulation of transcription

Solution2 : GCI:

molecular_function that regulates some transcription SubClassOf part_of some regulation of transcription

@balhoff would the GCI approach work with Arachne => entailment of involved_in given the property chain: enables o part_of -> involved_in in GPAD? If not , I guess we'd need some way of instantiating the relevant part_of links in the ontology for release.

[balhoff]

It would help to see this more concretely in an OWL file. But offhand, the right-hand side of the GCI falls outside of OWL RL I think. But if it influences the class hierarchy pre-inferred by ELK, it may still usefully feed into Arachne reasoning. But can you create example files for these?

[pgaudet]

@ukemi @balhoff
This deals with reasoning rather than MF refactoring. Is this being dealt with elsewhere ?

Thanks, Pascale

[balhoff]

@pgaudet over the past several months I've implemented a whole new reasoner to help handle this. 😄 In the future it may be integrated into Noctua. But I don't think we're using 'involved in regulation of' as a GPAD relation anymore, right? If @ukemi thinks there is an issue here for GPAD export, we should move to the Minerva tracker, otherwise I think we can close it.

[ukemi]

I have been using the explicit descriptions of regulation by creating the models where a function is part of a process that regulates another process/function. By using this method, if we know the mechanism of the regulation, we can assert it and the reasoner will still conclude it is a regulatory process. We also get the correct annotation to a regulatory process back. @vanaukenk I think that's what we decided. You could also make a model where the function is part of a regulatory process that regulates the downstream process. That might be clearer to annotators, but it's not as powerful because it doesn't allow you to specify a mechanism. I'd like to avoid involved_in_regulation_of. In my opinion it is a misleading shortcut relation that should result in an annotation to the regulatory process itself.