Open dustine32 opened 1 year ago
@thomaspd @huaiyumi This also means that every HUMAN gene in PANTHER 15.0 should have at least three "annotations" (they could include these "unknown [aspect]" annots), right? So there will be some genes that only have three "unknown" annotations if there are no IBAs to that gene.
From 2022-12-12 group meeting: The slim_terms
field value for any UNKNOWN
annotation should be the same UNKNOWN
term. For example:
{
"gene": "UniProtKB:12345",
"gene_symbol": "XYZ",
"gene_name": "Fake gene not real",
"term": "UNKNOWN:003",
"slim_terms": [
"UNKNOWN:003"
],
"qualifiers": "is_active_in",
"evidence": []
}
Oh good, this issue is still open! Found a new bug:
Genes that only have NOT
annotations for a certain aspect are not getting the appropriate UNKNOWN
annotation for that aspect. Ex:
O14531 NOT|enables GO:0004157
O14531 NOT|involved_in GO:0006208
Gene O14531
only has one annotation and it's to UNKNOWN CC
, the only aspect that doesn't have a NOT
. Those NOT
's above should truly be ignored and O14531
should have three UNKNOWN
annotations, one for each aspect.
In the annotations JSON input file, create an annotation for a gene and specific aspect if the gene has no other annotations to that aspect.
For example, if gene XYZ has annotations to GO:0000093 (a BP) and GO:0000095 (MF) but no CC annotation, a new annotation should be created like:
In the accompanying ontology JSON, these
UNKNOWN
terms should be defined:Some questions for @thomaspd @huaiyumi:
slim_terms
be empty in this case? If no, we will use the newOTHER
terms from issue #3.qualifiers
be blank or should I follow the GO default relations rules (MF=enables, BP=involved_in, CC=is_active_in)?evidence
contain anything like a gene ID or reference? Or just be empty?