Open dustine32 opened 3 months ago
@ukemi @deustp01 Plugging in this change and running the tests, our transport test example reaction (R-HSA-201669 in model R-HSA-201681) has changed because it's curated to a GO BP term. Showing the reaction before my change: And after: As you can see, the location info has been discarded along with the inferred "transport activity" function term. Is this ok with you?
Hi @dustine32. Yes and no. This is something we should discuss on a weeds call. For nuclear import, we have been inferring a transporter activity and hand waving that the enabler would be the nuclear pore. We need to discuss whether this is ok. If not, we have a biological process called 'import into the nucleus'. I think that process is a better asserted match for the black box reaction below that the regulation BP. I will put it on the agenda for a future call.
I think that process is a better asserted match for the black box reaction below that the regulation BP.
Questions, perhaps to set up that weeds discussion
You're not missing anything. I would argue for the import process assertion. I think we are seeing over and over that the inferred transporter activities are often problematic. It was a nice idea, but there are lots of exceptions where what is represented is actually a process. I still think we should discuss it with the other curators.
I still think we should discuss it with the other curators.
Definitely - exactly as for the transcriptional regulation we want to come up with a broadly useful annotation standard.
Here is how the reaction R-HSA-163743 "Transcriptional activation of ACACA by ChREBP:MLX" (the middle box) is currently converted:
There is no GO MF term or controller so the activity is Molecular Event. Reactome has curated a GO BP term so it is added and connected to the reaction via a "part of" relation.
The new conversion requirements should instead result in this pattern: The Molecular Event activity is gone with the Reactome-curated GO BP term now anchoring the "has input" and "has output" edges and connected to the preceding reaction via a "part of" relation.
In other words:
This rule should apply when a reaction class is Molecular Event (no GO MF term) and has a curated GO BP term. As discussed on the 2024-06-13 Reactome weeds call (@ukemi, @deustp01, @vanaukenk can confirm), a curated BP term should also override any inference of transporter activity in the conversion.
The code should decide which relation to connect between the reaction's curated BP node and its preceding activity node by fetching the preceding node's MF term and checking if the BP term is in the MF term's "part of" closure (ancestor terms). If yes, the relation will be "part of," otherwise the relation will be "causally upstream of".