Question about how annotation date updates are assigned in Noctua

[vanaukenk]

During last night's (2023-11-09) Noctua maintenance outage, I noticed many more models with that day's date than we typically have for an outage.

Some of these models were from the automated ontology updates and, based on the terms in them, it looks like other models might be from manual curator review and updates.

There was no comment for an automated change in those latter models.

The only date change I could see for the latter models was at the model level and not at the level of an actual 'annotation' when I looked at the models in the graph editor or the standard annotations in the Annotation preview.

I made some changes in a test model to see what might be happening and it looks like if I change anything about the evidence, the date on the 'annotation' updates, but if I replace one ontology term with another, e.g. a GO BP term, the date on the 'annotation' doesn't change.

I would have expected any change to the 'annotation' to trigger a date change for that 'annotation' in the Annotation preview or GPAD, but that doesn't happen.

Is this expected behavior wrt dates?

Thanks in advance for any insight!

@balhoff @kltm

[kltm]

@vanaukenk Do you have examples for some of these models?

I also recall that we made a change so that some actions do not trigger date update effects. I'll try and track that down...

[balhoff]

I think there are a couple of issues here. One is that the relationship between a node and its class (GO term) has no OWL annotations; no place to store date. The other is that to output a date for an annotation, the code has to select from a number of possible date values (on each node, on the edge, on the evidence). What happens now is that the date is taken from the evidence, since different evidences will result in separate annotation outputs. https://github.com/geneontology/minerva/blob/a6a9d0b84aac558fb077b3f2596bb5412710e74c/minerva-converter/src/main/resources/org/geneontology/minerva/legacy/sparql/gpad-relation-evidence-multiple.rq#L33

[kltm]

From software call discussion, it is apparent that date changes only propagate to the GPAD output when the evidence individual is manipulated. This is a little unintuitive, as things may be seen to me more recently updated in Noctua (versus the GPAD output). From @balhoff, as the GPAD is generated from a SPARQL query, the dates of the elements can be "maxed" to find the latest date...maybe.