1 September 2020 PAINT Conference Call
Annotation errors filtered - what is the desired behavior
Annotations with (a) the wrong aspect for a qualifier and (b) that fail taxon constraints are currently filtered out of PAINT.
Is this the behavior that we want ?
New species to consider from multi org group (Debby and maybe Val)
Mismatches Reference proteomes and MOD taxon IDs
Paul generated a spreadsheet with species with > 100 EXP annotations https://docs.google.com/spreadsheets/d/1jN2hrTkGIRNqOb1hSCX4MGSuzzXydS7C-YhUzXb01X4/edit#gid=0
I’ve highlighted in red the ones where the taxon IDs do not match between PANTHER/UniProt and GO (four total), and in yellow the ones where there is a distinct gene/protein ID space used in the GAF that we might want to double check to make sure it’s in the UniProt mapping file so we can ensure that we can match those up. I assume the standard MOD IDs will be in the mapping files, so I didn’t highlight those. I’ve also noted several other cases, where there are additional experimental GO annotations but to a distinct taxon ID. Not sure how we want to handle this.
- For pombe, UniProt Reference Proteomes uses taxon:284812, while pombase uses taxon:4896 (see email exchange 'pombe taxon').
- For Aspergillus fumigatus, the GAF is submitted by AspGD using taxID 746128, and UniProt has mapped those annotations to sequences with taxID 330879 (which NCBI shows as a child strain of 746128), so this one is similar to the S. pombe case in that UniProt uses the strain level rather than the species.
- For Leishmania major, the GAF is submitted by GeneDB using taxID 347515, and UniProt has mapped those annotations to sequences with taxID 5664 (which NCBI shows as a parent species to strain 347515), so this one is opposite to the S. pombe case in that UniProt uses the species level rather than strain.
- Maria Martin also write about S. cerevisiae: This is also true for S.cerevisiae where UniProt uses the reference strain S288c (taxon:559292), and there even less of the annotations are actually coming from experiments in that particular strain, yet UniProt (and SGD) annotate everything in GO to that taxon.
- WRT species, this is what UniProt/GOA does (from Alex):
Hi Huaiyu, We are using mapping provided by aspergillusgenome.org here http://www.aspergillusgenome.org/download/External_id_mappings/ If it has mapping to UniProt accessions, which belong to strain taxID, that what is used. We don't do mappings ourself we relaying on the DB authority to assign mapping to UniProt. I guess they are doing their best to do so. In the QuickGO fro example we include descendants by default if you filter by taxID. https://www.ebi.ac.uk/QuickGO/annotations?taxonId=746128&taxonUsage=descendants I'm not sure, but perhaps similar strategy can be applied to PAINT annotations. Best regards, Alex
Follow up on June action items (last call)
Update on Taxon constraints
- Dustin prepares a table at every release
- ACTION: Dustin to check with Jim on where are the 'full' taxon constraints are, so that we use the same.
- Jim will add the "missing" taxon constraints to the GO ontology release - https://github.com/geneontology/go-ontology/issues/19759
- Pascale would like the 'PAINT curation status' to be updated more regularly (at every monthly release?)
- ACTION: Anushya will do it.