Infer locations for activities based on their actors about pathways2go HOT 3 CLOSED

The reasoning in @goodb comment on 4 Sep sounds risky. Will this approach correctly handle cases where an activity (enzyme) is located in a membrane and accessible to substrates in an adjoining lumen, so the reaction actually occurs at the membrane - lumen interface and substrates / inputs / left bucket and products / output / right bucket are all in the lumen? And the cases where a protein with transport activity moves physical entities between the two cell components separated by that membrane?
Probably the answer, at least for Reactome, is that in all these cases the various reaction participants are all correctly assigned to compartments (failure to do that is a major curation error and we have QA that reliably finds such cases) so you should never need to use reasoning tools to assign cell component annotations.

from pathways2go.

@deustp01 just to clarify, we can't generate a MF occurs_in CC relation assertion directly from what I see in the BioPAX as the CC annotations only hang on the physical entities. The above logic (or adaptations you can suggest) is needed to produce those relations based on the participant CC annotations. I think we are safe here in that the rule will only activate if all of the participants are annotated to the same CC. And, if it goes haywire and we end up getting reactions that occur in multiple locations at once, the reasoner raises an alarm.

from pathways2go.

Agree - we should mostly be OK and the reasoner will give an additional level of QA when we make a mistake.

from pathways2go.