search

geneontology / noctua-form

2 stars 0 forks source link

Evidence lost for Function after adding an input. #162

Closed ukemi closed 1 year ago

ukemi commented 2 years ago

This morning in Noctua-dev I crated an model called Cblif (MGI:1202394) using the form editor. I made an MF annotation in the model with evidence. I saved the model and changed the title and state. Then I opened the model to add an input for the function and added evidence. When I saved the model, the evidence for the gene product enabling the function was lost.

ukemi commented 2 years ago

I just tried to add evidence manually to the function and I was unable to add it.

ukemi commented 2 years ago

I was able to add it back using the graph editor.

ukemi commented 2 years ago

We also noticed this during our training today and it is a serious issue. We tried to add evidence to an existing MF annotation and the original evidence was lost. Since this bug loses data, I am labeling it high priority.

vanaukenk commented 2 years ago

I also experienced this today when trying to add evidence for an 'occurs in' extension to an existing, evidenced MF. I lost all of the original evidence I had on the existing MF.

vanaukenk commented 2 years ago

@tmushayahama I tested this issue on noctua-dev this morning and it does not appear to be fixed, i.e. adding a 'has input' extension in the Noctua form table to an existing MF annotation removes the existing evidence from the 'enables by' assertion. Let's look at this again together. Current testing model on noctua-dev: http://noctua-dev.berkeleybop.org/workbench/noctua-form/?model_id=gomodel%3A627ec3ab00000007

vanaukenk commented 2 years ago

@tmushayahama will work on this in the coming week

ukemi commented 2 years ago

@tmushayahama Testing again on Noctua-dev.

  1. Create a model in the form: Enabled by Ihh Mmus
  2. I added a protein input to the receptor ligand activity and lost my evidence on the process.
ukemi commented 2 years ago

NB. This time the evidence was lost on the process this time, not the function. This is slightly different than before but is still bad.

image

vanaukenk commented 2 years ago

We tested again on the 2022-06-30 workbenches call. This is still an issue, with some new behavior from when the issue was originally reported. Here is the model we used for testing: http://noctua-dev.berkeleybop.org/workbench/noctua-form/?model_id=gomodel%3A62b627a100000059

Adding a protein as 'has input' to an MF removes evidence from the 'part of' BP associated with that MF. Later, after a screen refresh, two pieces of evidence now appear on the BP.

Adding a chemical as 'has input' to an MF removes the MF evidence, initially put duplicate evidence on the BP, but then both of the BP evidences disappeared.

vanaukenk commented 2 years ago

@tmushayahama - this may have to do with the order of operations, i.e. add vs remove.

vanaukenk commented 1 year ago

Fix is on @tmushayahama noctua instance. Will move to noctua-dev, test again there, and try to get out to production by next Tuesday.

vanaukenk commented 1 year ago

I've tested this issue on noctua-dev with model: http://noctua-dev.berkeleybop.org/workbench/noctua-form/?model_id=gomodel%3A62d7566e00000014

All seems to be working fine.

vanaukenk commented 1 year ago

Tested again on production - working as expected.