Odd non filtering cases

[ValWood]

https://www.pombase.org/gene/SPBC11G11.04 has a prediction for GO:0012505 | endomembrane system | IEA with ARBA00028953 but it is annotated to GO:0005801 | cis-Golgi network | IC with TRAPP complex which has this parent (also applies to sec22)

also rpn9 https://www.pombase.org/gene/SPAC607.05 GO:0000502 | proteasome complex | IEA with ARBA00029017 but it is annotated to GO:0008541 | proteasome regulatory particle, lid subcomplex | IDA which has this parent. (also rpn12, rpn7)

I can't see any obvious reason whey these are not filtered (I.e recent ontology changes, or odd relationships)? could you have a look to see if there is anything obvious?

[ValWood]

Another one mrps16 https://www.pombase.org/gene/SPBC354.06

has IEA to cytoplasm, but his is in the ancestry

https://www.ebi.ac.uk/QuickGO/GTerm?id=GO:0005763

[ValWood]

https://www.pombase.org/gene/SPCC5E4.07 rpl2801 seems to be no longer annotate to cytoplasm, even though it is in the parentage https://www.ebi.ac.uk/QuickGO/GTerm?id=GO:0022626

[ValWood]

These are all component terms. I wonder if there have been some recent GO changes that we are picking up, but have not yet filtered through to QuickGO?

@pgaudet. Does this make any sense to you?

In summary, we do not display IEAs if we already have annotation, but these IEAs are not filtered, implying some change in the ontology parentage?

[ValWood]

cox1 is picking up mitochondrion, even though we have mitochondrial inner membrane?

[kimrutherford]

It seems to have gone wrong when we switched to using the daily ontology build. I haven't tracked down why yet.

[kimrutherford]

I think I've tracked the what and where of this issue but I'm unsure about the why part. :-)

It looks like there has been a change to the relation definitions in go-basic.obo. In the last release the part_of Typedef looked like:

[Typedef]
id: part_of
name: part of
namespace: external
xref: BFO:0000050
is_transitive: true

But in the daily build and the snapshot, the is_transitive: true is not there anymore. That seems to be what is causing the changes in PomBase.

We use a handy program called relation-graph kindly created for us by Jim Balhoff (see https://github.com/geneontology/go-ontology/issues/17171#issuecomment-702852561) to create the transitive closure for storing in Chado. relation-graph uses the relations from go-basic.obo which probably explains the issue.

I tracked down the change in go-basic.obo to this commit:

commit 38898e10e25f4b0ac5e540fdd4bf25f6cc581601
Author: Jim Balhoff <jim@balhoff.org>
Date:   Tue Sep 10 12:42:19 2024 -0400

    Remove logical axioms from RO relations; these should be provided by the RO import. (#28891)

https://github.com/geneontology/go-ontology/commit/38898e10e25f4b0ac5e540fdd4bf25f6cc581601

@balhoff could you help us with this? What's the best fix?

[balhoff]

@kimrutherford very sorry about this! Trying to clean things up but not seeing all the ramifications. I put some of them back here: https://github.com/geneontology/go-ontology/pull/28995

[kimrutherford]

No problem Jim. We expect a few bumps in the road using the daily build of the ontology. :-)

I put some of them back here:

Thanks! I'll watch out for those to appear back in the daily ontology build.

Val, I patched the Typedef definitions in a local copy of go-basic.obo and ran an out-of-band Chado and website update. So I think the problems you found are fixed.

[ValWood]

Thanks for the speed. It wasn't major, but nice to fix.

[kimrutherford]

I'll watch out for those to appear back in the daily ontology build.

The fix has come through to ontology-build.geneontology.org/ so we're back to daily updates.

Thanks Jim!