What should be done with obsolete classes?
Closed this issue · 5 comments
Obsolete classes, and specifically obsolete phenotypes, are present as nodes in the graph.
Should they be removed after the fact?
Good question: I think they are noise and should be removed. I would be interested to know though if you find any cases where an obsolete class is connected to another node..
I found >14K edges involving obsoletes, though about 4600 of those are IAO:0100001 "term replaced by" edges - those probably aren't helping much.
Yeah this can be quite misleading I think.. I would drop all obsolete classes.
On a related note, KGX nearly has support for throwing out obsolete classes using node filters - I say nearly, because if I am reading the code correctly here, nodes can only be filtered by "category".
I wonder if Sierra could comment on how hard it would be to add support for filtering on arbitrary columns/node properties
The fix implemented here should be propagated as needed to other KG-Hub repos