Property or class? cheminfo ontology.

[Stian Soiland-Reyes]

Hi,

In

http://semanticscience.org/resource/CHEMINF_000465

I see:

  <rdf:Description rdf:about="http://semanticscience.org/resource/CHEMINF_000465">
    <rdf:type rdf:resource="http://www.w3.org/2002/07/owl#Class"/>
    <rdfs:label>ChemSpider validated synonym</rdfs:label>
    <rdfs:isDefinedBy rdf:resource="http://semanticscience.org/ontology/cheminf.owl"/>
    <rdfs:subClassOf rdf:resource="http://semanticscience.org/resource/CHEMINF_000044"/>
    <dc:identifier>CHEMINF:000465</dc:identifier>
    <ns0:IAO_0000115>A preferred name in ChemSpider by virtue of having been validated by a curator.</ns0:IAO_0000115>
    <ns0:IAO_0000117>Person: Colin Batchelor</ns0:IAO_0000117>
  </rdf:Description>

This is also reflected in http://semanticscience.org/ontology/cheminf.owl

How is this owl:Class meant to be used?


In Open PHACTS we see this kind of data from RCS:

from http://data.openphacts.org/1.4/rdf/OCRS/20131111/CHEMBL/SYNONYMS_CHEMBL20131111.ttl.gz (~300 MB)

@prefix cheminf: <http://semanticscience.org/resource/> .
@prefix ops: <http://ops.rsc.org/> .
@prefix owl: <http://www.w3.org/2002/07/owl#> .

cheminf:CHEMINF_000465 a owl:AnnotationProperty . # Validated Synonym
<http://ops.rsc.org/OPS1769270> cheminf:CHEMINF_000465 "2-naphthalenecarboxamide, 4-chlorodecahydro-4a-hydroxy-N-(4-methoxyphenyl)-8a-methyl-3-oxo-, (4R,4aS,8aS)-"@en .


I know reusing a term both as an owl:Class and an owl:AnnotationProperty this is possible with OWL 2 Punning - but it sounds like something is not quite right in this case. Did RCS get it wrong? I would expect a 'preferred name' to be a property, not a class - why is it defined as a class in the ontology if it is really meant to be used as a property? Is there a rdf:value or similar property that should be used to provide the actual literal synonym? If so, with which property should we relate ops:OPS1769270 to the synonym instance?

I tried to visualize the cheminf ontology at http://www.essepuntato.it/lode/http://semanticscience.org/ontology/cheminf.owl  but I must admit I get a bit confused by semi-numeric term names like "iao0000027" and "CHEMINF_000466".



--
Stian Soiland-Reyes, myGrid team
School of Computer Science
The University of Manchester
http://soiland-reyes.com/stian/work/ http://orcid.org/0000-0001-9842-9718

[Colin Batchelor]

Hello Stian,

(Already answered off-list but I'll reply here.)

This is intentional, as it's a label and not meant to be inferred over, hence treating it as an AnnotationProperty but I've asked Andrea about what he's actually doing which might require a different approach.

Best wishes,
Colin.

[Egon Willighagen]

Hi Stian,

We recently had this discussion too, and then it was said to be
intended use from a CHEMINF perspective...

On Wed, Jan 7, 2015 at 12:14 PM, Stian Soiland-Reyes
<st...@mygrid.org.uk> wrote:
> I would expect a 'preferred
> name' to be a property, not a class - why is it defined as a class in the
> ontology if it is really meant to be used as a property?

I guess this is the design decision of CHEMINF: it captures
"descriptors" of things (chemicals) and models these as OWL
ontologies. I do not see why "preferred name"s should be any different
from other things describing that chemical.

That is besides the question about whether this can be used as a
property, which was new to me last summer too.

> Is there a rdf:value or similar property that should be used to provide the actual
> literal synonym? If so, with which property should we relate ops:OPS1769270
> to the synonym instance?

Do any of these examples come close to what you mean?

https://code.google.com/p/semanticchemistry/wiki/Examples

> I tried to visualize the cheminf ontology at
> http://www.essepuntato.it/lode/http://semanticscience.org/ontology/cheminf.owl
> but I must admit I get a bit confused by semi-numeric term names like
> "iao0000027" and "CHEMINF_000466".

Yeah, that is a decision design in line with many other ontologies.
The downside is clear, but it has the advantage that you can fix the
title/label without getting conflicts with the URI... there may be
other advantages I forgot...

But they are hard to remember and the reason why I added a few tables
to the above Examples page...

Egon


--
E.L. Willighagen
Department of Bioinformatics - BiGCaT
Maastricht University (http://www.bigcat.unimaas.nl/)
Homepage: http://egonw.github.com/
LinkedIn: http://se.linkedin.com/in/egonw
Blog: http://chem-bla-ics.blogspot.com/
PubList: http://www.citeulike.org/user/egonw/tag/papers
ORCID: 0000-0001-7542-0286
ImpactStory: https://impactstory.org/EgonWillighagen