Open suzialeksander opened 1 year ago
Note that this is the opposite of what we do for the Reactome models. For those, we infer and retain the causal relationship between the functions.
@ukemi Right! In fact, the last commit 35f360c was made to fix a Reactome test that broke due to this. The fix ties the causal relation removal to the use of common intermediate input/output molecules as in the image from the top comment: If relation chain 1 exists, then relation 2 will be removed.
For Reactome, (as you said above) I haven't yet switched the input/output molecule handling over to the scheme shown for relation chain 1, but weren't we talking about doing that? If yes, is the desired Reactome behavior that we have both relation chain 1 and relation 2 between input/output-connected reactions?
I'd actually like both. To me removal of the causal relation (even if it can be inferred) goes against the GO-CAM data model which is supposed to link functions with causal relations..
I think that the group decided that we would follow the method described above. We need to do this with the Reactome import as well.
follow the method described above
Sorry - which method above? Keep both of links 1 and 2 or only keep 1?
There are some redundant edges added to models upon import:
Example from YeastPathways_YEAST-SALV-PYRMID-DNTP
I think the other major relation that did this was "casually upstream of".
Can we remove these before https://github.com/geneontology/pathways2GO/issues/275?
Also tagging @thomaspd