Has the ontology snapshot changed? missing regulation relationships

[ValWood]

Describe the issue/bug

We used to get regulation inferences from the ontology shapshot, but this no longer seems to be the case.

Expected behavior

Here is a good example why they are needed for inference.

The translation repressor sum2 https://www.pombase.org/gene/SPBC800.09 no longer gets mapped to "cytoplasmic translation" by inference.

Without the regulation relationships I don't see a way to annotate this to cytoplasmic translation?

A clear and concise description of what you expected to happen.

To Reproduce

Please include any information you can about: what, where, when, and why. Please include any URLs involved in the issue.

Screenshots

If applicable, add screenshots to help explain your problem.

Additional context

Add any other context about the issue that you feel may be important. E.g. browser version, time of day.

[kltm]

As far as I know, there have been no major changes to the ontology build, but @balhoff may be able to add additional information here.

[ValWood]

It's a bit weird.

GO:1904689 | negative regulation of cytoplasmic translational initiation does not propagate to GO:0002181 cytoplasmic translation but GO:2000766 negative regulation of cytoplasmic translation does

[ValWood]

It might be a problem at our end. I will reopen if necessary....

[balhoff]

I don't know of any intentional changes.

[ValWood]

so none of these are annotated to cytoplasmic translation by propagation (ancestry): (I expect them to slim to cytoplasmic translation)

https://www.pombase.org/gene/SPAC27E2.02 IMPACT homolog, cytoplasmic translational regulator Yih1 is annotated to GO:1990611 | regulation of cytoplasmic translational initiation in response to stress

https://www.pombase.org/gene/SPAC688.14 ribosome L32 lysine methyltransferase Set13 is annotated to negative regulation of cytoplasmic translational initiation in response to stress

https://www.pombase.org/gene/SPBC800.09 translation initiation inhibitor (predicted) is annotated to negative regulation of cytoplasmic translational initiation

Maybe these should be part_of rather than regulation of? @pgaudet what do you think?

but in this case what could be annotated to these terms?

[ValWood]

This issue is also related

https://github.com/geneontology/go-ontology/issues/21557

[ValWood]

Maybe it is an issue at our end after all. Sum2 is annotated to cytoplasmic translation in AmiGO. I'll ask @kimrutherford to take a look.

[mah11]

https://github.com/geneontology/go-ontology/issues/21557#issuecomment-854485669

[suzialeksander]

@ValWood were you able to locate the cause of this mapping/ancestry issue- is it the change in GO's reasoning re: "regulates a whole process" vs "regulates part of the process"?

[ValWood]

Hi no I did not look at this yet, but I think it is due to the recent changes.

This is how it looks:

translation --regulation of translation (regulates) ----regulation of translation initiation (is_a)

and translation --translation initation (is_a) ----regulation of translation initiation (regulates)

MF translation repressor has an is_a connection to regulation of translation initiation

So yes the recent changes (removing propagation over some relationships) prevented the annotation inheritance, it seems.

However, the proposed more recent change to prevent inheritance over ALL regulation relationships, will negate any fix applied now, this issue will return if/when this proposal is implemented.

It will be odd that that common initiation factors like GO:0004694 eukaryotic translation initiation factor 2alpha kinase activity will no longer be in the translation bin. It is quite a big change and might be confusing for users. Are we OK with this?

Is the MF-BP relationship correct? Can we have a template that shows how these factors should be annotated (regulation/not regulation)? i.e Maybe initiation should be defined as beginning with "eukaryotic translation initiation factor 2alpha kinase activity phosphorylation of the initiation factor? In these cases, if we really do expect regulators to be in a process bin do we need to define the process 'start' to include them?

In this scenario, the regulation connection here would be between the activity and target entity (I'm also making an assumption here that Noctua exports these connections as "blah" rather than "regulation of blah" otherwise we are in a pickle...(can anyone confirm? all my attempts to export a simple model to check these things so far have failed)

Or, maybe I just need to get used to the annotation not propagating into the translation class, but definitely, we would need to enrich over "regulation" to get the best results if that is the case.

@thomaspd @cmungall @pgaudet @RLovering @vanaukenk

[hattrill]

Seems to be a lot of points here, so not sure where to drive in and if I have the correct take on this! I would class the initiation factors as core components of translation and therefore should be annotated to 'translation initiation'. Anything that acts to inhibit initiation or to regulate the activty of the initiation factors to the regulation terms.

In terms of what you get when you do a query - it should be absolutely clear to the user that there are options to have the 'core' process components and/or the regulators (that act on the core factors or direct regulators of core factors).

[ValWood]

but initation factors also include regulators. It isn't very clear to me what is classed as core and what is regulation. Everyone seems to have a different take on this.

[hattrill]

So it's the MF-BP inferrence here that is causing the issue. IMO - something can have MF 'regulator activity' but still be part of the process. Making the hard link in the ontology does not seem correct here in this example. .

[ValWood]

Yes, that is the problem (well part of it!)