Transitive closure for over regulates for MF by default can be confusing in some cases

[ValWood]

De we want to include regulates_closure for molecular function?

This give weird results. For instance if I search for all of the gene products with "protein kinase activity" for pombe, the list includes protein kinase activators and inhibitors like cig2 and rum1.

I think most people would find this misleading.

Val

[kltm]

This is most likely mostly a question for @cmungall (and @hdietze), as the best place to make any change along these lines would probably be during loading.

[cmungall]

The general framework here would be to have the default closure field depend on the focus class. This will be required in any case for #201

E.g.

• c in MF
  • Question: which gene products enable c?
  • isa_partof_closure
  • Why: enables o SubClassOf -> enables
• c in BP
  • Question: which gene products are involved in c?
  • regulates_closure
  • Why: enables o (part_of|regulates) -> involved_in
• c in CHEBI
  • Question: which gene products are involved in a process that have c as participant?
  • involved_in_o_has_participant_closure (see #201)
  • Why: enables o (part_of|regulates) o has_participant -> involved_in_o_has_participant

etc

Of course, the boundary between MF and BP can be arbitrary. Question for @ValWood -- what would you want the default behavior for protein phosphorylation to be? On the one hand, it's the exact same thing as with protein kinase acticity. However, there is an implicit difference in question, and I think GOC members would want kinase gene sets to exclude regulatory and phosphorylation gene sets to include regulatory. Is this just a GOC expectation? What would a wider cross-section of biologists think? Something for @dosumis to think about in the MF refactoring. Whichever way we model things, AmiGO provides the framework to answer the questions in a sound fashion.

[kltm]

267 seems to be a reformulation.

[cmungall]

@kltm correct, looks like I repeated myself: https://github.com/geneontology/amigo/issues/267#issuecomment-152879983

My fault, should not be using the tracker as a substitute for proper requirements documentation.

TL;DR: choose the default closure type based on the type of term. We should also do more to makethe reasoning more transparent and intuitive.