richard apodaca wrote:
> --- On Tue, 9/29/09, Nina Jeliazkova <nina at acad.bg> wrote:
>
>   
>> Sorry, the correct link to implemented services  is
>> http://opentox.org/toxicity-prediction/opentoxservices
>>     
>>> New version, currently under revision
>>> http://opentox.org/dev/apis/api-1.1/?searchterm=API
>>>
>>> One note - if the information above is hard to find
>>>       
>> and not obvious
>>     
>>> where to look for, we indeed would need to reorganize
>>>       
>> at least the front
>>     
>>> page of OpenTox site.
>>>       
>
> Thanks for the link. But it looks like a list of services.
>   
Can you clarify , why do you think so?
> What I'm actually after is a list of resources, in the REST sense of the word - a domain model if you will.
>
> For example, a "resource" in this context might be "Prediction".
>   
Well, not intending to go into terminology discussion, but  Model or
Algorithm are just resources.  Models can be applied on a dataset or
substance (this is defined in the API).

I would personally interpret prediction as something that already
happened, after applying a Model, otherwise it sounds more like an
"activity" than a "resource".
> When we talk about making a Prediction, a lot of other concepts go with that. For example, we might perform a Prediction on a Substance or a Dataset. 
We apply "Models" on a Substance of a Dataset.  Models are created by
applying Algorithms on Datasets.
> That Dataset might be composed of Items, for example. The Substance may in turn be composed of Components. It's these domain concepts and how they're aggregated that make a domain model. Only with a domain model in hand can you start thinking about resources.
>   
Datasets are composed of compounds, and might have descriptors
(features) defined.  We agreed on a flexible approach, where any
compound property (descriptor, name, endpoint) is a feature. Feature can
have meta information, assigning specific meaning (e.g. descriptors).
> The point is not to get bogged down in the details, but rather to create a common language for toxicity prediction. A language that can be understood by domain experts (Toxicologists, in vivo pharmacologists) and developers alike.
>
> So far, I haven't found anything like this for OpenTox. If there is no such thing, how interested would the OpenTox group be in developing one?
>   
It seems to me you are confused by the terms, adopted in OpenTox, which
are more or less following machine learning vocabulary. Otherwise
OpenTox resources quite closely follow your suggestions. Does my
explanations help?

Best regards,
Nina
> Best,
> Rich
>
> ___________________________________
>
> Richard L. Apodaca
>
> http://depth-first.com      Blog
> http://metamolecular.com    Company
>   
> _______________________________________________
> Development mailing list
> Development at opentox.org
> http://www.opentox.org/mailman/listinfo/development
>