Dear all, 
 As it is substantial for the implementation of QPRF reporting services
to extend the ontology API (taking into account the specification of the
JRC at http://ecb.jrc.it/qsar/qsar-tools/qrf/QPRF_version_1.02.pdf , I
think that the following are needed:

1. The class 'ot:Report' is subclassed by 'ot:QPRFReport' and
'ot:QMRFReport'. Question: Are there any other reports needed there?
Maybe 'ot:ValidationReport'?

2. As you read in the updated version of the ontology I sent yesterday,
a new ontological class called 'ot:AdequacyInfo' is needed to include
the information of section 4.x of the QPRF specifications (see link
above). Four dartatype properties are related with this class. These are
'ot:adequacyConclusion', 'ot:adequacyOutcome', 'ot:regulatoryPurpose'
and 'ot:regulatoryInterpretation'. 

3. I don't think that any modification is needed for section 1 of the
specifications. We just extend the domain of the object property
'ot:dataEntry', to contain 'ot:QPRFReport' as well. Then, for example,
we may have triples like:
http://<server>/report/qprf/1 ot:dataEntry X
X ot:compound http://<server>/compound/1
Note: Only one compound should be related to a QPRF report (JRC
specifications).

4. Section 3.1. : What I suggest here is the the endpoint/predicted
variable is correlated to the QPRFreport using the property
ot:predictedVariables. For example:
http://<server>/report/qprf/1 ot:predictedVariables
http://<server>/feature/1

5. Section 3.2.a. Make use of ot:algorithm

6. Section 3.2.b. Introduce the property ot:qmrfReport which may
correlate a Model with a QMRFReport. For example:
http://<server>/report/qprf/1 ot:model http://<sever>/model/1
http://<sever>/model/1 ot:qmrfReport http://<server>/report/qmrf/1
http://<sever>/model/1 ot:qmrfReport http://<server>/report/qmrf/2
http://<sever>/model/1 ot:qmrfReport http://<server>/report/qmrf/3
...

7. About section 3.3. : In every model, one can find a link to an
applicability domain model (or at least an AD algorithm). So in the RDF
representation of the qprf report one finds the model used to make the
prediction and the AD model/algorithm for that model. Now we need to
handle sections 3.3.a (domains) and 3.3.c (3.3 Considerations on
structural analogues) [We'll discuss 3.3.b later]. So I suggest to
introduce the class ot:DomainInfo for 3.3.a. Once ot:domainInfo is
introduce we can have the properties:
* ot:descriptorDomain : Model U algorithm --> String
* ot:structFragmentDomain : Model U algorithm --> String
* ot:MechanismDomain : Model U algorithm --> String
* ot:MetabolicDomain : Model U algorithm --> String

8. Section 3.3.b. (Structural Analogues): This is a specific n-ary
transitive relation between compounds. For that reason (thinking
SQL-wise) we introduce the generic class ot:Relation which is subclassed
by ot:CompoundRelation which is subclassed by ot:StructuralRelation. A
membership in a relation is declared using the property
ot:relationElement (see updated ontology). As far as 3.3.c. is concerned
we can use rdfs:comment for it, which is something like a comment on the
applicability domain. For example:
X rdf:type ot:StructuralRelation
X rdfs:comment "the user provides a discussion on structural analogues".

9. About 3.2.c. : I suggest that the predicted value is contained in the
RDf representation of the QPRF report explicitly to avoid posting a
compound to the descriptor calculation proxy service and then to model
service. 3.2.d. can be handled using rdfs:comment.

Any comments or ideas?
Best regards,
Pantelis


On Mon, 2010-06-14 at 17:18 +0300, chung wrote:
> Dear all,
>  I send you again an updated version of the OpenTox ontology. Please
> check out if the attached file opens with Protege because there have
> been some problems (possible Protege version compatibility problems). I
> think we should finalize the error reporting part of the ontology, so
> please take a look at ot:ErrorReport and ot:errorReportProperty.
> 
> Best regards,
> Pantelis.
> _______________________________________________
> Development mailing list
> Development at opentox.org
> http://www.opentox.org/mailman/listinfo/development