Closed lpalbou closed 3 years ago
For reference, the model I was testing: http://noctua-dev.berkeleybop.org/editor/graph/gomodel:5b91dbd100002241
I am closing this ticket for now as after looking more closely, the GOLr queries were responsible for the side effect (e.g. finding biomacromolecule when looking for a compound).
The ticket can also be deleted as my explanation was both incomplete and wrong (e.g. both compound & GP have a chemical entity root, so the way to differentiate them is based on the biomacromolecule term).
I am creating this ticket here as I believe it's a GOLr issue @kltm @cmungall .
While doing testing for ART, I was surprised I could change a GP by a compound:
Oxaloacetic acid reference
The way it works, @tmushayahama retrieves the closure of the term entered in the "find" box to constrain the results of the autocomplete in the replace box. In essence, we only want to allow the change of a GP by another GP, but not a GP to a compound for instance.
Interestingly, after clicking change, the enabled_by line of ART for that activity disappears (as it should) which means the closure retrieved when filling the table is correct. I think the one used for that table comes from minerva/bbop and neo, while the first query used in the find/replace box on the left directly targets GOLr (golr.berkeleybop.org) - @tmushayahama unsure why we are even querying GOlr directly and that's probably something to change.
Long story short, I checked the GOlr query made for a GP (daf-4) and filtered it for readability.
Part of the response showing isa_partof closure:
CHEBI tree for that compound:
And indeed, it seems puzzling that GOLr would indicate in the isa_partof closure that daf-4 is not only a molecular entity but also a chemical entity.. So we may have a bug somewhere in the index creation of that instance of GOLr golr.berkeleybop.org.
Digging a little bit more, turns out that gene daf-4 would have the same number of terms for each isa, isa_partof and regulates_closure, which i find suspicious too: