Ola,


On Thu, Aug 26, 2010 at 7:58 PM, Ola Spjuth <ola.spjuth at farmbio.uu.se>wrote:

> Nina,
>
> Thanks for your feedback. Yes, we can run a local application server, that
> should not be a problem. If the war is all we need to deploy then it should
> not be too difficult. Is the requirement for MySQL hardcoded or is it easy
> to switch to e.g. HSQLDB?
>

It's hardcoded, sorry.  Not sure how easy is to switch all of the
functionality, because of some MySQL specifics as bitsets.  Some time ago
there was a request to support Oracle, but this was not explored because of
lack of resources. If anybody is interested to introduce HSQLDB /other DB
support, would be nice to join efforts.


> Do you use any object-relational mapping?
>

No (for performance reasons).  JDBC resultsets are iterated over and
serialized in the requested mime type.


> Is it easy to turn on/off different components (for example if we would
> like to start with just supporting predictions)?
>

All services will be on, but no problem to use predictions only.


>
> It would be nice if we could deploy our own OpenTox reference application
> (if I understand things correctly this is what Ambit is)
>

Ambit is a project older than OpenTox, which has a Swing standalone
applicationinitially (AmbitXT for download),  with functionality similar to
Bioclipse, but with a MySQL backend.  It was extended during OpenTox to
expose all its functionality (and more) as OpenTox REST services.

Other OpenTox partners have their own implementation of (subsets) of
services , in Java and Ruby as well.



> as a war, and package our own models as additional wars (according to some
> spec). Is this the intention to build? If so, what is the extension
> mechanism for this? Is there any documentation for how such plugins are
> structured? We are building something similar based on the OSGi component
> model and the Eclipse framework, but it would be great if we also could
> deploy our models as OpenTox services.
>
>
Extensions are not something that is supported at the moment, but ambit code
is maven based and dependencies are quite well separated.  We can discuss
further what's the best way towards an extension mechanism.  For example
Toxtree loads plugins via Java classloader, but this is not currently used
in the rest services.  Separate wars will result in separate applications,
while if different services are packaged as jars, this could be the same
application.

Best regards,
Nina



> Cheers,
>
> /Ola
>
>
> On 26 aug 2010, at 13.29, Nina Jeliazkova wrote:
>
> Ola,
>
> In general OpenTox models, as everything else in OpenTox should be consumed
> via REST interface.  However, one can run REST services locally.  Our
> implementation (the one that is running at
> http://apps.ideaconsult.net:8080/ambit2 ) could be downloaded from
> http://ambit.sourceforge.net/downloads.html and launched inside your own
> instance of Tomcat (or other servlet container).
> It covers almost all OpenTox API (dataset, algorithms, models), besides
> validation and reporting. Besides dataset services, that allows to import
> compounds and data and do queries via identifiers or substructrue or
> similarity, the war contains all Toxtree models , as well as many CDK
> descriptors  and some Weka algorithms that can be used to build models.
>
> The ambit download page contains links to the source code of the services,
> as well as two projects to illustrate a REST client to services, as well as
> a demo project showing how to create your own model service, but this is
> pretty much work in progress.
>
> We'll be working to improve documentation as well.
>
> Best regards,
> Nina
>
> On Thu, Aug 26, 2010 at 1:25 PM, Barry Hardy <
> barry.hardy at douglasconnect.com> wrote:
>
>>  Hello Ola:
>> Good to see your post on the OT dev list and I will let response happen
>> there with regards to inspecting and downloading models etc.
>>
>> I also thought it could help to point out:
>> Some specific people to potentially interact with directly on OpenTox with
>> regards to current tasks/activity areas:
>> - Nina (e.g., data services, ToxTree, ToxPredict,...)
>> - Christoph (ToxCreate, Model Creation...)
>> - Martin (Validation services...)
>> - Tobias, Pantelis (Algorithms, descriptor calculations...)
>> - Andreas Maunz who is currently working on authorisation and
>> authentication
>> - Andreas Karwath who is currently working on reporting
>> - Vedrin on testing, server development, deployment...
>> - Roman who has been recently assigned the task, supported by the
>> developers, to work on improving documentation, guidance, tutorials etc.
>>
>> @All - I am scheduling a VM with Ola, Lars and Egon for 15.00 CET next
>> Weds Sept 1 to discuss Bioclipse-OpenTox developments.  Let me know if you
>> can join.
>>
>> Barry
>>
>> -------- Original-Nachricht --------  Betreff: [OTDev] Some questions on
>> REST/interoperability  Datum: Thu, 26 Aug 2010 11:35:07 +0200  Von: Ola
>> Spjuth <ola.spjuth at farmbio.uu.se> <ola.spjuth at farmbio.uu.se>  Antwort an:
>> opentox development mailing list <development at opentox.org><development at opentox.org>  An:
>> opentox development mailing list <development at opentox.org><development at opentox.org>
>>
>> Hi,
>>
>> I am investigating OpenTox on behalf of the Bioclipse project (http://www.bioclipse.net). Egon Willighagen has already created means in Bioclipse to query and retrieve data using a REST interface, I wish to ask more about models and predictions.
>>
>> I (together with collaborators) have developed some predictive models based on public toxicity data, and wish to investigate possible interoperability with the OpenTox community in terms of publishing and using models. We try to run as much as possible offline; is it possible to download and run OpenTox models offline or is it mandatory to go via a REST interface? Do people share models (e.g. as war archives) so others can deploy them locally (or mirror them)? I am also interested in the feasibility of bundling existing OpenTox datasets and predictive models as OSGi services in addition to REST services.
>>
>> Is there a simple tutorial for how to consume an OpenTox service to predict an endpoint for a single chemical structure? In true REST manner, can a simple URL accomplish this (I could not find any examples on the website)? Is there a reference Java implementations for consuming OpenTox REST services, especially predicting endpoints for chemical structures?
>>
>> I am impressed with the OpenTox project, it seems like a well designed project! However, it is a cumbersome process to get up to speed with the current status/design so forgive me if my questions are trivial or already answered on the web page.
>>
>> Best regards,
>>
>> /Ola
>>
>> --
>> Ola Spjuth, PhD
>> Department of Pharmaceutical Biosciences
>> Box 541, Uppsala University
>> Sweden
>>
>> Email: ola.spjuth at farmbio.uu.se
>> Web: http://www.farmbio.uu.se
>>
>> Bioclipse: http://www.bioclipse.net
>>
>> _______________________________________________
>> Development mailing listDevelopment at opentox.orghttp://www.opentox.org/mailman/listinfo/development
>>
>>
>
>
> --
>
> Dr. Nina Jeliazkova
> Technical Manager
> 4 A.Kanchev str.
> IdeaConsult Ltd.
> 1000 Sofia, Bulgaria
> Phone: +359 886 802011
>
>
>
>


-- 

Dr. Nina Jeliazkova
Technical Manager
4 A.Kanchev str.
IdeaConsult Ltd.
1000 Sofia, Bulgaria
Phone: +359 886 802011