OLS ontologies with issues

[jamesamcl]

These are ontologies in the OLS config which we have issues loading (NOT those imported from OBO, there is a separate issue for OBO ontologies at #1).

ontology id	purl	problem	fix
FIXED pending config update - sdgio	http://purl.unep.org/sdg/sdgio.owl	not found	change purl to https://raw.githubusercontent.com/SDG-InterfaceOntology/sdgio/master/sdgio.owl
FIXED pending config update - teddy	http://www.biomodels.net/teddy/TEDDY	redirects to html at https://sourceforge.net/projects/teddyontology/	copy owl file to cluster
cco	http://www.bio.ntnu.no/ontology/CCO/cco.owl	java.lang.NullPointerException: Cannot read field "properties" because "this.ontologyNode" is null
rexo	http://www.bio.ntnu.no/ontology/ReXO/rexo.rdf	java.lang.NullPointerException: Cannot read field "properties" because "this.ontologyNode" is null
gexo	http://www.bio.ntnu.no/ontology/GeXO/gexo.rdf	java.lang.NullPointerException
atol	http://www.atol-ontology.com/public/telechargement/atol.owl	javax.net.ssl.SSLHandshakeException: PKIX path building failed: sun.security.provider.certpath.SunCertPathBuilderException: unable to find valid certification path to requested target
eol	http://www.atol-ontology.com/public/telechargement/eol.owl	javax.net.ssl.SSLHandshakeException: PKIX path building failed: sun.security.provider.certpath.SunCertPathBuilderException: unable to find valid certification path to requested target
lbo	http://data.bioontology.org/ontologies/LBO/download?apikey=8b5b7825-538d-40e0-9e9e-5ab9274a9aeb	obo format
probonto	https://raw.githubusercontent.com/probonto/ontology/master/probonto4ols.owl	Turtle but not in content-type
pride	https://raw.githubusercontent.com/PRIDE-Utilities/pride-ontology/master/pride_cv.obo	obo format
unimod	http://www.unimod.org/obo/unimod.obo	obo format
afo	http://afo-ols.semanticsfirst.com/ontologies/afo	dead link; domain does not resolve
hpath	https://raw.githubusercontent.com/Novartis/hpath/master/src/hpath.obo	obo format
vido	https://raw.githubusercontent.com/infectious-disease-ontology-extensions/ido-virus/master/ontology/vido.owl	owl xml

[jamesamcl]

(1) The OLS4 dataloader is an RDF tool and therefore only supports loading RDF files. This means that other non-RDF OWL serialisations such as OBO format and OWL XML are never going to be supported (though of course they can be converted prior to loading). For these (very few) cases we can either ask the upstream ontology vendors to provide an RDF/XML file, or possibly outsource conversion to Robot.

(2) Though we support all the different RDF serializations, the majority of the ontologies are provided without any content-type or any useful file extension to indicate which serialization format they contain. For example, this ontology from the OLS config is Turtle, but the file extension is owl and the content-type is text/plain. No suggestion of Turtle encoding anywhere.

Even the OBO foundry ontologies do this. If we resolve for example http://purl.obolibrary.org/obo/ro.owl it redirects to https://raw.githubusercontent.com/oborel/obo-relations/master/ro.owl. File extension: .owl, content-type is text/plain. While the file content is RDF/XML, there is nothing to suggest that it isn't, for example, OWL XML, or Turtle, or JSON-LD. We only know how to load it in OLS4 because RDF/XML is the hardcoded default.

Why does this work in Protegé and OLS3? Because OWLAPI literally bruteforce loads ontology files by trying every loader until it finds one which works.

While we could probably do something similar in OLS4, I think ultimately it is up to the ontology developers to provide correct metadata, if not by content-type then at least by file extension. The whole .owl thing is a mess. If it's RDF/XML it should be .xml and if it's Turtle it should be .ttl. OR if it really wants to be .owl it should be served up with a content-type.

So TL;DR I think we should continue to default to trying to read .owl files as whatever content-type is provided and falling back on RDF/XML.

[jamesamcl]

Here's one done right: https://www.w3.org/ns/prov-o-20130430

While there's no file extension, the content-type is text/turtle; no guessing game required!

[lschriml]

Checking out the new OLS4.
Looks great.
Noticed that my Human Disease Ontology is not in the list of ontologies.
Please let me know if there is an issue, that we can address.
Cheers,
Lynn

[bgyori]

Hello, I just noticed that ProbOnto is not in OLS4 but was in OLS3 and then found this issue where it is listed in the table above. Unfortunately ProbOnto's own ontology browser hasn't been working for a long time now (https://sites.google.com/site/probonto/ontology) so getting it working in OLS4 would be nice. Would this be possible? Thanks!

[jamesamcl]

Hi @bgyori ,

We currently attempt to load it from https://raw.githubusercontent.com/probonto/ontology/master/probonto4ols.owl which is a Turtle serialization with the file extension .owl.

OLS4 is much stricter about loading ontologies than OLS3: we use Apache Jena instead of OWLAPI, so it doesn't bruteforce different serializations, instead relying on the file extension and/or content-type, with RDF/XML as the default when such information is unavailable.

Ideally the maintainers would provide an RDF/XML version, or else rename the file to .ttl. If the ontology is no longer maintained (I see the last commit on https://github.com/probonto/ontology was 6 years ago) we could do a one-time conversion and store the converted version on the EBI server perhaps.

[bgyori]

If the extension is the only issue, I could pull request the same file with a ttl extension into the probonto repo and see if they will merge it. Would that be sufficient to solve this?

[jamesamcl]

Yes that should work! Meanwhile I copied it onto EBI NFS for the next load.

[serjoshua]

Refer to #453 for updated information regarding failing OLS ontologies.