search

pantherdb / fullgo_paint_update

Update of Panther and PAINT DBs with monthly GO release data
0 stars 0 forks source link

Obsolete redundant IBDs on same node #35

Open dustine32 opened 5 years ago

dustine32 commented 5 years ago

From https://github.com/geneontology/go-annotation/issues/2660: image We recently introduced a script to obsolete redundant IBDs that are on different nodes in a tree. For example, an IBD on AN1 to GO:0015225 already accomplishes what another GO:0015225 IBD to AN1-descendant AN2 would do. Therefore, the GO:0015225 IBD to AN2 will be obsoleted. As it turns out, this meant redundant IBDs that are on the same node together were ignored and left alone. Will need to fix this script to include this scenario.

Also, pay attention to different with lists for the IBDs (e.g. [SPAC1B3.15c, SPAC1B3.16c] vs. [SPAC1B3.16c]). Does a distinct with list make an IBD worthy of saving?

pgaudet commented 5 years ago

following

pgaudet commented 5 years ago

Also, pay attention to different with lists for the IBDs (e.g. [SPAC1B3.15c, SPAC1B3.16c] vs. [SPAC1B3.16c]). Does a distinct with list make an IBD worthy of saving?

How come there are different lists anyway ? These should all be part of the same evidence.

dustine32 commented 5 years ago

@pgaudet Agreed! Though doesn't the PAINT tool give the curator some options for which GO annotation evidence to associate with an IBD? I could be confused about this.

mugitty commented 5 years ago

When an IBD annotation is created by the PAINT tool, the tool determines the list of all experimental annotations that can be used for the IBD. It then groups them based on the qualifier. If there is more than 1 qualifier, the system prompts the curator to select the qualifier. Based on the qualifier selected by the curator, the PAINT tool only uses those experimental annotations that have the qualifier selected by the curator.

dustine32 commented 5 years ago

Ah, that's where I was mistaken. Thanks @mugitty !

pgaudet commented 5 years ago

However, if there is a NOT annotation on the leaf, the positive annotation should NOT be propagated.

Right ?

mugitty commented 5 years ago

Any leaf that that has experimental evidence that is used for creating the IBD will not have the propagated annotation.

pgaudet commented 5 years ago

Right ! So I don't understand why we are seeing those. There any many examples.

mugitty commented 5 years ago

Do you have an example of it appearing on PAINT?

pgaudet commented 5 years ago

In PAINT you dont see the positive annotation. It seems like a pipeline/export problem.

dustine32 commented 5 years ago

@pgaudet @mugitty I believe by design the leaf SPAC1B3.15c would still get a positive GO:0015225 IBA if the positive GO:0015225 IBD using SPAC1B3.16c (vht1) as evidence is still valid AND there's no IKR/IRD on or above leaf SPAC1B3.15c in PAINT to prevent the positive IBA propagation.

Though, those redundant positive IBAs on SPAC1B3.15c are indeed a pipeline/export problem that I'm working to fix right now.