Make (and use) isoform promotion GORULE

[kltm]

As part of @sierra-moxon 's work, we explicitly made the choice to have the following data transform: when there is an isoform available in the GPI, use that as the GPAD bioentity, as it is the most specific (and can be recovered from the GPI).

We would like to make sure that this transform is well-documented with a GORULE (in addition to the mention in the spec), as it is an active transform of data.

Tagging @pgaudet

[cmungall]

Can you expand on this? When exactly would this rule be applied?

I think I'm missing some context. The isoform ID should only be used if there is evidence for function in an isoform.

[kltm]

The GORULE portion is fungible, but I think it would be good to make explicit any transformations that are being done, if not immediately obvious (although, as pointed out elsewhere, this is implied in the spec).

IIRC, the decision was reached in a managers' call, which seems to be poorly documented ("MGI remainders" in):
https://wiki.geneontology.org/Projects_update_meeting_2024-03-13
https://wiki.geneontology.org/Projects_update_meeting_2024-02-28

I think the summary given by @sierra-moxon is best here: geneontology/gopreprocess#36

[pgaudet]

Hi @cmungall

It seems the documentation is not super clear. GPAD 2. doc is here: https://github.com/geneontology/go-annotation/blob/master/specs/gpad-gpi-2-0.md

I had understood that since GPAD2.0 does not have the equivalent of Column 17, column 1 contained the 'active' entity (although now I don't see that written out clearly). My understanding was that GPAD had to be used in conjunction with the GPI, which specified the 'Parent_Protein' (GPI column 8) that corresponds to what is currently in GAF column 2. Otherwise, there is no way to capture isoform-specific data.

So, this is why @sierra-moxon took what was was in GAF column 17 and moved it to GPAD column 1. However I now realize that the GPI never describes the type of the 'Parent protein', which may results in entity type confusion.

Pascale

[cmungall]

On Fri, Apr 12, 2024 at 12:11 AM pgaudet ***@***.***> wrote: Hi @cmungall <https://github.com/cmungall> It seems the documentation is not super clear. GPAD 2. doc is here: https://github.com/geneontology/go-annotation/blob/master/specs/gpad-gpi-2-0.md you’re right, I thought there was more guidance on this I had understood that since GPAD2.0 does not have the equivalent of Column 17, column 1 contained the 'active' entity (although now I don't see that written out clearly).

My understanding was that it was fine for the gpad to have a gene id in col 1, unless the intent is to capture isoform specific function

My understanding was that GPAD *had* to be used in conjunction with the GPI, which specified the 'Parent_Protein' (GPI column 8) that corresponds to what is currently in GAF column 2. Otherwise, there is no way to capture isoform-specific data. So, this is why @sierra-moxon <https://github.com/sierra-moxon> took what was was in GAF column 17 and moved it to GPAD column 1.

But this is fine, and does not require a gpi lookup (the reverse direction of course requires a gpi lookup)

…

However I now realize that the GPI never describes the type of the 'Parent protein', which may results in entity type confusion. Pascale — Reply to this email directly, view it on GitHub <#2287 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAAMMOLZLOGVJ6F5M24RYI3Y46CLLAVCNFSM6AAAAABFWJU3V2VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDANJRGE2DCOJZGM> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[sierra-moxon]

I believe this is complete; on snapshot now