Recent PRO ids do not display in the annotation table

[hdrabkin]

Two PRO ids just new in Neo (as of last night), can be used for annotation in Form and Graph (autocomplete). However, after saving annotation the table does not display the gene product.

Use of a old PRO id does not have this issue.
Use of a regular gene product id does not have this issue.

Question: time delay between appearance of PRO id in annotation drop down vs table display? If so, how long.

[kltm]

@hdrabkin Well, this likely due to the different frequencies that we're updating things. Noctua (minerva) itself is being updated (restarted) every other week, as it can be a little time consuming. To make sure that correct identifiers are in place as quickly as possible, NEO has been updated about once a week, on Fridays. Looking at that cadence, one can see where gaps can happen.
It fix this a little, I'm going to go ahead and move the NEO updates back a day (as they are very low risk) to Thursdays, so the restart can happen more or less with the NEO update on those weeks. Otherwise, I'm not sure that it's desirable to decrease the NEO update cadence, but that's something that @vanaukenk might have an opinion on or want to consider.

[kltm]

Noting that this runs into issues like geneontology/pipeline#35; possibly the addition of new entities would affect cadence as well.

[vanaukenk]

Thanks @kltm
We could ask curators if they'd be okay with less frequent NEO builds, albeit NEO builds otherwise synced to the Noctua maintenance outages.
Having an entity available in the autocomplete but not viewable in the Noctua form table is an understandable point of confusion.

[kltm]

There was a push at some point for a very fast cadence on NEO, but the details are slipping my mind at the moment. Assuming that a lower cadence is otherwise fine, I'd have no problem linking to to the outages (that and adding these to the main pipeline and the timing issues around those being deferred).

[vanaukenk]

It's still a trade-off. Rapid updates of NEO are desirable for responding to curators' needs, but I understand the source of confusion surrounding the table display.

Maybe there are other solutions for this particular problem, e.g. the table just displays whatever the enabler is, regardless of 'type'. I'd have to ask @tmushayahama about possible options.

[hdrabkin]

Probably it's only MGI that this effects: since we continuously add new annotatable objects (isoforms and proteoforms). I'll bet most DBs see no change in their GPI overlarge periods of time. However, for us, it is more important to be able to use the ids asap, since there is already a delay between submitting to pro, getting the new ids, getting the new ids to get put into our gpi (depends on when new file available on the pro site to pick up), then waiting up to a week to have them available in Noctua. This was just noticed because was first time using the form for PRO ids; they DO appear in the graph view after making the annotation in the form. Just wasn't expecting a difference. The check with export to gpad does show the expected annotations, as well asn the annotaton preview.

[tmushayahama]

@vanaukenk it will have a chain reaction of bugs if the type is not known. So I think it's better to have all types resolved in minerva. Examples include editing, ART (constraints), complexes. distinguishing between form types CC Only vs Activity Unit. CC Only.

[vanaukenk]

To summarize actions:

1. For now, we will continue with the weekly NEO build but that will now happen on Thursdays before the Noctua maintenance outages.
2. The Noctua maintenance outages will continue on their current 2nd/4th Thursday schedule.
3. Document the current schedule and behavior so curators understand what they might see (or not see) on the Noctua form table.

@hdrabkin @kltm @tmushayahama - please confirm. Thx.

[kltm]

@vanaukenk 1 and 2 correctly summarize the way operations are currently run.

[tmushayahama]

@vanaukenk We should bring back the shex error checks we disabled or the Noctua Form 1 table error display

[vanaukenk]

@hdrabkin
I've checked this model post-update and it looks like the isoform names are now there.
If all looks good to you, too, please feel free to close the ticket.