Closed ElaineMcA closed 5 years ago
@LucaFumis initial investigations:
Seems like it could be an issue with one of the OT util services - perhaps otUtils.getEcoLabel
The thing is in the response we pick “gene2variant”: { “functional_consequence”: “http://purl.obolibrary.org/obo/SO_0001583” and look it up in the response for a more human friendly description but there isn’t anything…
"evidence_codes_info": [
[
{
"eco_id": "GWAS",
"label": "Genome-wide association study evidence"
}
],
[
{
"eco_id": "cttv_mapping_pipeline",
"label": "CTTV-custom annotation pipeline"
}
],
[
{
"eco_id": "ECO_0000205",
"label": "curator inference"
}
]
],```
So in there we look for ‘SO_0001583’
but there isn’t anything.
Just to elaborate a little on the initial investigation (apologies my English was a little "rusty").
The function otUtils.getEcoLabel
is a convenience method that looks up a code in an array. It takes the following from the API response:
evidence.evidence_codes_info
that's the list of code-labels as per comment aboveevidence.gene2variant.functional_consequence
from which we extract the code at the end of the string, e.g. SO_0001583
in http://purl.obolibrary.org/obo/SO_0001583
If the code is not in the array, it just returns the code. This is what's happening here.
This might be a data thing too. I know there are fake ontology terms are passed straight to the UI from evidence strings. We should return both the ontology URI and the human readable label in the API and then the UI can display a hyperlink.
Training on 31Jan2019 - options (i) FE workaround (ii) amend training pending 19.02 release.
This needs fixing - Do we only get the labels from the evidence_codes_info array, or are we supposed to get them from the service i.e. http://www.ontobee.org/ontology/SO?iri=http://purl.obolibrary.org/obo/SO_0001583
If the former we must not be building this correctly
There is the ECO elasticsearch index, which is already used by the rest API elsewhere. I think the simplest solution is to use that to lookup labels instead of the hard-coded static mapping.
@iandunham @ElaineMcA @LucaFumis It will be addressed by this issue #388 and this PR
Now requires pipeline rerun to resolve - this will be coordinated to include fixes (i.e. #382 )ongoing post 18.12.
Just to add that we spotted this with the Genomics England tiering data as well.
QA approved 17Jan2019. Ready to push to production.
We no longer list "missense variant" etc, but rather its SO term e.g. https://www.targetvalidation.org/evidence/ENSG00000258839/EFO_0000756?view=sec:genetic_association Whereas for rare variants we have now a mix of SO term and trinucleotide expansion, although the latter has a SO term as well i.e. SO_0002165 e.g. https://www.targetvalidation.org/evidence/ENSG00000102081/Orphanet_908?view=sec:genetic_association
Although the users can hover and click on the '?' in the column name for the common diseases and be redirected to the /variants page where the user will find out what SO 0001583 means...
this functionality (of being redirected to /variants page and find out what the SO term is) is not available for rare diseases (or the rare variants section):
(reported by @deniseOme on 2nd Jan 2019)