search

EBISPOT / OLS

Ontology Lookup Service from SPOT at EBI
http://www.ebi.ac.uk/ols
Apache License 2.0
97 stars 41 forks source link

The PSI-Mass Spectrometry ontology hierarchy does not look right in OLS website #623

Open javizca opened 2 years ago

javizca commented 2 years ago

Dear colleagues,

https://www.ebi.ac.uk/ols/ontologies/ms

It seems that something is not right because in the first hierarchical level of the ontology, many terms from the ontology (which should not) appear and also the complete UNIT ontology as well. This is not right. I don't know whether it is the fault of the parser or we have formatted wrongly the OWL file in any of the recent changes.

Could you please help?

Thanks,

Juan

henrietteharmse commented 2 years ago

I think there are multiple issues here:

  1. I see the following in the ms.owl file: <owl:Class rdf:about="http://purl.obolibrary.org/obo/xsd_positiveInteger"/>. This to the best of my knowledge this is not the standard way to create an ontology.
  2. The ontology is inconsistent, which can also be an indication that the way the ontology is created may be problematic.

In both cases 1 and 2 I strongly suggest the creators of the ontology to get in touch with @dosumis to discuss ways to address this.

  1. The ontology is making use of punning, i.e., "acre" (UO_0010025) is both a class and an individual. This may be due to an issue in OLS, but this is something that is still under investigation.
henrietteharmse commented 2 years ago

OLS has been updated to fix (3) above. (1) and (2) still needs to be addressed by the designers of the PSIMass Spectrometry ontology.

javizca commented 2 years ago

No, the issues are not solved still. Work in progress.

henrietteharmse commented 2 years ago

Though we can update MS on OLS, it will still be broken on OLS while these are still an issue.

edeutsch commented 2 years ago

Okay, thank you, @dosumis how can we address these issues?

dosumis commented 2 years ago

HI @edeutsch - Maybe a quick call to talk through the issues would be the most efficient way to proceed? If you're interested, we could help you switch to a build pipeline with QC to prevent issues like this from occurring.

matentzn commented 2 years ago
  1. is fine.. Class-individual is legal punning (I would not generally recommend it though)

I don't see an OLS issue. OLS uses the MS PURL http://purl.obolibrary.org/obo/ms.owl to download the ontology. This ontology imports the entirety the PATO and STATO. This is why your top level looks so "untidy":

image

There are many unsats as @henrietteharmse says:

image

Open MS in Protege and debug it. Or use ROBOT explain:

robot explain -i ms.owl --reasoner ELK -M unsatisfiability --unsatisfiable random:10 --explanation unsat_1.md

This will help you identify the logical errors! Hope that helps!

dosumis commented 2 years ago

It seems that something is not right because in the first hierarchical level of the ontology, many terms from the ontology (which should not) appear and also the complete UNIT ontology as well.

To improve the view in OLS, you can add preferred roots to your OLS config. This will set the terms that are visible at the top level by default. @serjoshua should be able to point you to documentation on how to do this.

henrietteharmse commented 2 years ago

@dosumis Preferred roots will not sort out issues like the xsd_anyURI etc being displayed at the root. That should be resolved in the ontology?

image

matentzn commented 2 years ago

obo:xsd_anyUri is such a weird way to talk about xsd:anyUri datatype.. that said if the preferred root is some other class somewhere, no other classes will be shown regardless of if they are floating to the top, right? https://www.ebi.ac.uk/ols/ontologies/mondo has lots of floating classes but looks all nice and tidy..

henrietteharmse commented 2 years ago

@matentzn Agreed about obo:xsd_anyUri. That is why I suggested they get in touch with your teams.