Hi Ola,

On 3 August 2011 10:58, Ola Spjuth <ola.spjuth at farmbio.uu.se> wrote:

> Hi,
>
> I was wondering if there is a definition how results from predictions are
> to be structured in order for clients to interpret it.


Yes, all should follow ot:Dataset RDF format in opentox API ontology.
http://opentox.org/api/1.1/opentox.owl

N3 Example:

[]    a       ot:Dataset ;
      ot:dataEntry
              [ a       ot:DataEntry ;
                ot:compound <
http://apps.ideaconsult.net:8080/ambit2/compound/100/conformer/143935> ;
                ot:values
                        [ a       ot:FeatureValue ;
                          ot:feature <
http://apps.ideaconsult.net:8080/ambit2/feature/27457> ;
                          ot:value "10.335800170898438"^^xsd:double
                        ]
              ] .


<http://apps.ideaconsult.net:8080/ambit2/compound/100/conformer/143935>
      a       ot:Compound .

ot:NumericFeature
      a       owl:Class ;
      rdfs:subClassOf ot:Feature .

Where ot:DataEntry is roughly analogous to a row in a spreadsheet , and
ot:Feature - to a spreadsheet column, ot:FeatureValue - to a spreadsheet
cell.

This had been explained in various reports & publications, as well as on
opentox site.  (Yes, the documentation could be improved :)



> For example, if a result has a predicted toxicity value (e.g. Double)


All these are represented as ot:Feature, and values i as ot:FeatureValue
entries, as in the above example.

The feature should be instance of any of ot:NumericFeature, ot:StringFeature
, ot:NominalFeature


> with accuracy (Double)


The feature should be instance of ot:ModelConfidenceFeature and links to the
feature with predicted value via ot:confidenceOf property.
This is used by the validation service.


> and a subtructure (set of atoms),


Apart from agreeing to use additional ot:smarts property for structure
features , this was not standardised further.


> how this is defined for clients to reason and visualize it. I now see HTML
> code in some prediction results, and I'm sure this cannot be the intention.
>

No, it is not :)

Nina

[1] http://www.jcheminf.com/content/3/1/18

>
> Cheers,
>
> /Ola
> _______________________________________________
> Development mailing list
> Development at opentox.org
> http://www.opentox.org/mailman/listinfo/development
>