UniPathway resolves to a dead link

[cmungall]

This may be more an annotation issue than a site one

links such as:

go to:
http://www.grenoble.prabi.fr/obiwarehouse/unipathway/upa?upid=UPA00886

which says it is no longer supported. Note our link is correct, following the link from quickgo takes us to the same place.

What should we do here? Although this particular human annotation is superfluous, it seems that unipath annotations can often be very useful for bacteria?

Note that unirules sometimes have a more specific unipath ID than a GO process ID:
https://www.uniprot.org/unirule/UR000079203

But I believe we get the more granular annotation from the unipath2go mapping (which was last updated 2 years ago)

cc @pgaudet

[amorgat]

UniPathway is a dead project. Thanks to Murphy law, two years ago, the production and the backup servers crashed in one week. Unfortunately, we don't have resource to continue this project. You should remove the links.

[pgaudet]

@amorgat We can't just remove the links - if this is dead, we'd remove the entire annotation lines. Who generates these annotations? Some Uniprot pipeline ?

[pgaudet]

just to be more precise: without the link, we dont know where there IEA is from, so we cannot keep it.

[tonysawfordebi]

We still have a pipeline that creates these annotations, based on the last-known-good UniPathway2GO mapping, and UniPathway cross-references in UniProt entries.

[pgaudet]

Looks like there are 6,910,457 annotations with GO_REF:0000041 in UniProt .. that's quite a bit ! But if it cannot be mapped to an active resource, I'd be inclined to remove them.

Does Rhea generate GO annotations that could be used to replace these ?

[amorgat]

Definitely something to discuss together.
Would it be possible to add this item in the Monday/Tuesday agenda of our next meeting?

[ukemi]

I added it to the end of the day Tuesday because it seems like it fits into the data-flow/pipeline area.

[tonysawfordebi]

I've raised the issue here about whether we should continue to generate & publish UniPathway-based annotations.

[amorgat]

@ukemi thanks David

[pgaudet]

The plan discussed in Geneva - ontology editors meeting is to move the UniPathway to OLS and link there.

[pgaudet]

@kltm Do we plan to do this before the next GOC meeting ?

[kltm]

@pgaudet This looks like an ontology editors issue; I do not know the status.

[pgaudet]

@cmungall Will be up in OLS in a couple of days.
Need to make the link -need to edit the dbxref

[pgaudet]

@cmungall what would the URL be for UniPathway? I dont see it here:
https://www.ebi.ac.uk/ols/ontologies

[cmungall]

It looks like @amorgat registered unipathway with bioportal in 2018

https://bioportal.bioontology.org/ontologies/UPA

is this the most up to date one Anne?

[cmungall]

oh it seems something is wrong with the upload, I will contact them and cc you ann

[cmungall]

I forget it is registered with OBO

can be browsed in OLS

e.g URLs like

https://www.ebi.ac.uk/ols/ontologies/upa/terms?iri=http%3A%2F%2Fpurl.obolibrary.org%2Fobo%2FUPa_UPA00323

[amorgat]

@cmungall, as unipathway can be considered as a dead resource, I guess, I should remove it from bioportal, obo foundry, etc
What do you think?

[pgaudet]

From @alexsign : UniProt plans to retire UniPathway xrefs
the pathways IDs are available from OLS and from Bioportal, but any information about what composes the pathways is gone: see
https://bioportal.bioontology.org/ontologies/UPA
and
https://www.ebi.ac.uk/ols/ontologies/upa/terms?iri=http%3A%2F%2Fpurl.obolibrary.org%2Fobo%2FUPa_UPA00323

So, I will remove the links from the ontology.

Thanks, Pascale

[pgaudet]

@sandraorchard
We have an issue in that the annotations generated by the UniPathway pipeline are not traceable, since UniPathways are nowhere on the internet, and the last update (as far as I can reconstruct from comments in the present ticket) is from 2016 at the latest.

We'd like to remove annotations coming from that pipeline, unless you have another solution?

Thanks, Pascale

[alanbridge]

Hi Pascale,

please go ahead and remove the annotations from UniPathway.

All the best, Alan

[redaschi]

Hi Pascale,
I was not aware of this issue until this week. Given that it is open since 5y, can you maybe wait a bit longer before you remove the UniPathway2GO pipeline to let us assess the impact for users? At least in Swiss-Prot we would like to minimize the loss of these BP GO terms. My quick investigation shows we have

• 53,695 GO terms with IEA:UniProtKB-UniPathway

• 143,776 CC PATHWAY annotations

  • 111,658 with an evidence from a HAMAP-Rule or PROSITE-ProRule
    GO BP terms based on these should in theory either already exist with IEA:UniProtKB-UniRule, or be created after you stop the UniPathway2GO pipeline. If two different IEA pipelines predict the same term, what is your precedence rule? i.e. would we currently have IEA:UniProtKB-UniPathway or IEA:UniProtKB-UniRule if both predict the same GO term?
    BUT: We have to check whether these rules have the GO term from https://ftp.ebi.ac.uk/pub/databases/GO/goa/external2go/unipathway2go that corresponds to the pathway, or at least a related GO term (Murphy's law, in the first example I checked, MF_00627, the HAMAP curator used "GO:0006567 threonine catabolic process", while in unipathway2go there is the child term GO:0019518).
    478 annotations also have a PubMed, which we should put on the GO, see next section.
  • 14162 with an evidence not from a HAMAP-Rule or PROSITE-ProRule
    I think these are manually curated evidences (I see 8339 with PubMeds, incl. the 478 mentioned in previous section). We want to check whether the corresponding GO terms already have the PubMed or ISS or are IEA:UniProtKB-UniPathway. If the latter, we should find a way to add the manual evidences.
  • 17,964 without an evidence
    These are likely experimental, but from a time before curators added evidences, and I see zero chance for a manual retrofit, so we will likely loose the GO BP terms that were linked to these annotations.

Cheers, Nicole

[pgaudet]

Hi @redaschi

thanks for your feedback. This pipeline is entirely within UniProt, so you guys can decide when to switch that off, so problem.

@alexsign