Hi Christoph,
	OK, now it works more or less. Just a few comments. The posted
parameters "prediction_feature" and "dataset_uri" are not supposed to be
parameters of the model. Second, the parameters in your models appear to
not have a scope (optional/mandatory) - can you fix this. Finally, DC
properties are datatype properties - you might consider using other
OpenTox object properties like for example ot:hasSource.

Here's the output of Jena (as warnings) as it attempts to parse the
resource http://ot-test.in-silico.ch/model/255 :

Non-literal value found for property :
http://www.opentox.org/api/1.1#paramValue
URI of that value is : http://ot-test.in-silico.ch/algorithm/fminer/bbrc
Non-literal value found for property :
http://www.opentox.org/api/1.1#paramValue
URI of that value is : http://ot-test.in-silico.ch/dataset/1249
Non-literal value found for property :
http://www.opentox.org/api/1.1#paramValue
URI of that value is :
http://ot-test.in-silico.ch/dataset/1229/feature/Hamster%
20Carcinogenicity
Non-literal value found for property :
http://purl.org/dc/elements/1.1/creator
URI of that value is : http://ot-test.in-silico.ch/model/255


Best regards,
Pantelis Sopasakis


On Mon, 2011-08-22 at 18:43 +0200, Christoph Helma wrote:
> Dear Pantelis,
> 
> 
> > Some IST  models cannot be parsed.  Jena says:
> > SEVERE: qedit.task.ImportModel at ec1c4d failed:
> > com.hp.hpl.jena.ontology.ConversionException: Cannot convert node
> > "http://www.opentox.org/api/1.1#ModelPrediction"^^http://www.w3.org/2001/XMLSchema#string to OntClass: it does not have rdf:type owl:Class or equivalent
> > com.hp.hpl.jena.ontology.ConversionException: Cannot convert node
> > "http://www.opentox.org/api/1.1#ModelPrediction"^^http://www.w3.org/2001/XMLSchema#string to OntClass: it does not have rdf:type owl:Class or equivalent
> > 
> > which means that http://www.opentox.org/api/1.1#ModelPrediction should
> > be an Ontological Class (or at least a resource) but it's a literal -
> > the error sources from
> > http://toxcreate3.in-silico.ch/model/9304/predicted/value (see
> > attachment).
> 
> Thanks for alerting us!  I hope to have fixed it in our development
> branch. It will take some time until the new version makes it through
> our tests to the master branch and to "offical" models. If you want to
> test in the meantime you can use one of the models at
> ot-test.in-silico.ch/model (after tomorrow).
> 
> > Question: Shouldn't all feature URIs follow the pattern /feature/{id}?
> 
> As we do not have a separate feature store, features are kept under the
> URI of the parent object (model or dataset as appropriate). I remember
> faintly that we have discussed that some time ago on the mailing list
> and we agreed that feature URIs as sub-URIs of other services are ok.
> 
> > Also I think that the triple:
> > 
> > http://toxcreate3.in-silico.ch/model/9304/predicted/value
> > http://www.w3.org/2001/XMLSchema#anyURI
> > http://toxcreate3.in-silico.ch/model/9304/predicted/value
> > 
> > is not much meaningful.
> 
> Agreed, but ist valid OWL-DL ;-). Does it cause any problems - if not I
> would prefer to keep it at the moment, because I fear to break other
> parts of OWL-DL serialisation.
> 
> > 
> > @All:
> > Please use http://www.mygrid.org.uk/OWL/Validator to make sure that your
> > representations are OWL-DL valid and http://www.w3.org/RDF/Validator/ or
> > http://inspector.sindice.com/index.jsp to check the structure of your
> > RDFs. 
> 
> We do nightly tests with http://www.mygrid.org.uk/OWL/Validator (if it
> is not down), but somehow prediction features slipped through. Sorry,
> OWL-DL validation of prediction features should be in our tests by now.
> 
> Best regards,
> Christoph
>