Integrating JChem in a workflow?

Hi to all,


I was wondering if anyone (ChemAxon developers or else...) is thinking of integrating JChem in a (free and open source) workflow; I am beginning to work with

Kepler, built on top of PtolemyII - it has got some interesting modules and can be developed with Eclipse


KNIME, based on Eclipse, with very useful modules already - data minig (WEKA), statistics (R), some chemoinformatics from the CDK, and Tripos has begun to add some extensions (Schrodinger is in the forum, but I did not see anything yet...)

Any ideas or interests?


Regards,


Nicolas

HI Nicolas,





I moved this topic here, hope you don't mind.





Yes we are looking at KNIME right now but it hasn't been resolved, are there any products in particular you want to see there?





Alex

Not at all, I was not sure where to put it... ;-)

Back to the question: it would be nice to have the usual chemoinformatics tools like fingerprint(s) generation, clustering, converting, potential database retrieval and insertion... basically all the JChem Base suite + the calculator plugins!

Cheers Nicholas,





That is interesting but how about Marvin - is this something you see a big need for?


Alex

Hey Alex,

Quote:

how about Marvin - is this something you see a big need for?

Sure, actually - It would be great to visualize results from a(ny) workflow analysis...


Is this something that you really plan to do?


Let's say I would try to write new plugins in a workflow (KNIME and/or Kepler), would you guys have anything against it (providing, of course, we make the whole thing available to the community...)?


Cheers,


Nicolas

Hi Nicholas,





Thanks for the information.

Quote:

Let's say I would try to write new plugins in a workflow (KNIME and/or Kepler), would you guys have anything against it (providing, of course, we make the whole thing available to the community...)?

Good thinking and best intentions - I have not looked at Kepler but we would need to speak to KNIME about this, there may be commercial issues - let me check and come back to you.





Cheers / Alex

Here is what I got from KNIME, since they already got some plugins developed from Tripos (and maybe Schrodinger?)

Quote:

The deal with Tripos and Schrödinger is simple. They get to do their own integration and give back to the KNIME development in exchange.

About Kepler, I contacted people from the Resurgence project...


Cheers,


Nicolas

HI Nicholas,





The feedback from KNIME is

Quote:

People are free to program their own nodes. But they can't give out nodes that call commercial tools. So if Nicolas wants to do this for his own use, no problem. From a KNIME point of view, no special license keys are needed.

For KNIME my feeling is that if you make the connector available to universities they can use the connector and ChemAxon tools for free. I will correct this post if I am wrong.





Good works Nicholas


Cheers / Alex

HI again Nicholas,





KNIME confirm that connectors can be used with any free software - limitations only occur when the connector is being used with a paid for software - so connecting to our toolkits (since you are free academic user) will be fine





Alex

Hi Alex,


Thanks for the details. Hopefully I will be able to implement a few things in the near future!


To be continued...


Nicolas

Good works


Alex

Hi Alexa and saettel,


I want to know have you integrated JChem in a workflow? with KNIME


if yes is the workflow available for academic use?


Thanks

HI pushkar,





We have not implemented any functionality in KNIME as yet, we have a couple of test pieces but these are not finished yet. Can you tell me which functionality you would like to see?





What ever we have will be made available for free use by the Academic Package people





Cheers/Alex

Alex - you started already? Would it be possible, by any chance, to have a preview of the pieces you wrote? If you wrote it as an Eclipse/KNIME plugin, it should already be possible to play around and it would give me a start to see if I can come up with somethings of my own (not that I pretend to be a developer, but if I can contribute ;-)


Pushkar (& Alex) - I have a virtual meeting tonight (San Diego + Zurich + Caen-Normandy...) with people from the Resurgence project in Kepler, I will keep you posted!


Nicolas

Hi Nicholas,





Nice :)





I asked the developers to respond about what's already available in KNIME but this is very early and limited coverage





About Kepler: We have not started anything here (even a commercial review) and I am not sure about how relevant this would be to other users, perhaps you can comment. Another issue is to confirm provision conditions (is there any cost associated with using or distributing connectors - either academic or commercial), perhaps you could pass on my email address (aa__at__chemaxon__.__hu) and ask they get in touch (I could see no contact point at their web pages)





Cheers/Alex

Hi Nicolas,





I've created two knime test nodes from MolImporter, and MarvinSpace, but they are just very basic nodes (just a few hours' work).


They were created for test purposes only - to see if they can be integrated at all.


I can give you the eclipse/knime source and some explanation if you'd like to.





Judit

Hi Judit,


I would very much like that, as I said, it could give me a head start on my own development...


Sources and explanation are very welcome, if you could be so kind...


Cheers,


Nicolas

All right, let's see.


I think the best start is to read the Technical Report of Knime, especially the II. part (Architecture).





After installing knime, I've created a sample workflow to get more familiar with it. Their Quickstart guide was a great help, it took just some minutes to see it work.


To create the nodes I've used the knime extension howto, that provided help too, but I felt it a bit too linear and limited to the sample they provide.





I've put the sources to http://www.chemaxon.com/shared/knime, please bear in mind that it is just a test code, far from being complete.


Some notes:


- Legal note: these sources cannot be used for commercial purposes.


- In mspace some code change was also necessary, and since it is not released yet, I've uploaded the test version of the necessary libraries too.


- It was not obvious for me, that the main configuration is done in the plugin.xml files, and they are displayed with tabs.





The workflow


Data flowing between nodes is wrapped in a DataTable. If I understood well, numeric and string values can be easily used here.


I decided to create a workflow of 2 nodes: one imports molecules from file, puts them in the datatable rows in string (smiles) format, and the other node finally displays them.


Note, that I didn't want to figure out a more sophisticated data flowing now.





The MolImporter node


A configuration dialog was created (MolImporterDialog) where input file can be set, and the DataTable was built (in MolImporterNodeModel).





The MarvinSpace node


Note, that MarvinSpace can read molecules from files directly, so the first node was necessary only to test the data flow.


Configuration dialog was disabled here.


The main things: the mspace canvas is created in the MarvinSpaceNodeView constructor, data changing is handled in MarvinSpaceNodeView.modelChanged().





This might be enough for a start, please give a feedback soon.


Have a good work!


Judit

Thanks a lot, Judit.


I got the files, figuring the different parts now - comparing with what's done in the tutorial from the new KNIME Node-Extension and trying to reproduce what you have done.


I guess it is not enough to put everything in the workspace (in your case, C:/Applications/eclipse/workspace/ judging from your .classpath file) and start eclipse, or do you have to import from the package explorer view (I guess plugins and fragments)? Forgive my naive question, I am very new to all this - frantically reading/learning about java and eclipse!


To be continued...


Nicolas

Hi Nicolas,





Just copy them to your workspace, create new Java Projects (e.g. from Package Explorer) with the exact names, and the wizard will detect the sources.





Cheers,


Judit

All right, here is the feedback!


First I ran into problems since I was using KNIME 1.2.0 and you developed it with 1.1.2. Since the declarations are not the same, it did not work.


Using the eclipse developer version 1.1.2, I was able to import your java projects as you said in the Package Explorer view (I previously tweaked the .classpath to point to my workspace)


I got a few warnings, like in MolImporter

Code:

Severity and Description   Path   Resource   Location   Creation Time   Id The field MolImporterNodeDialog.m_dialogPanel is never read locally   MolImporter/src/chemaxon   MolImporterNodeDialog.java   line 37   1172161920953   7 The method createFileNamePanel() from the type MolImporterNodeDialog is never used locally   MolImporter/src/chemaxon   MolImporterNodeDialog.java   line 63   1172161920953   8 The method createPreviewPanel() from the type MolImporterNodeDialog is never used locally   MolImporter/src/chemaxon   MolImporterNodeDialog.java   line 118   1172161920953   9

and in MarvinSpace

Code:

Severity and Description   Path   Resource   Location   Creation Time   Id The import java.awt.Container is never used   MarvinSpace/src/chemaxon/marvinspace   MarvinSpaceNodeView.java   line 3   1172161945468   10 The import org.knime.core.data.DataCell is never used   MarvinSpace/src/chemaxon/marvinspace   MarvinSpaceNodeModel.java   line 6   1172161945500   13 The import org.knime.core.data.DataColumnSpec is never used   MarvinSpace/src/chemaxon/marvinspace   MarvinSpaceNodeModel.java   line 7   1172161945500   14 The import org.knime.core.data.DataColumnSpecCreator is never used   MarvinSpace/src/chemaxon/marvinspace   MarvinSpaceNodeModel.java   line 8   1172161945500   15 The import org.knime.core.data.DataRow is never used   MarvinSpace/src/chemaxon/marvinspace   MarvinSpaceNodeModel.java   line 9   1172161945500   16 The import org.knime.core.data.def.DefaultRow is never used   MarvinSpace/src/chemaxon/marvinspace   MarvinSpaceNodeModel.java   line 14   1172161945500   18 The import org.knime.core.data.def.DoubleCell is never used   MarvinSpace/src/chemaxon/marvinspace   MarvinSpaceNodeModel.java   line 15   1172161945500   19 The import org.knime.core.data.def.IntCell is never used   MarvinSpace/src/chemaxon/marvinspace   MarvinSpaceNodeModel.java   line 16   1172161945500   20 The import org.knime.core.data.def.StringCell is never used   MarvinSpace/src/chemaxon/marvinspace   MarvinSpaceNodeModel.java   line 17   1172161945500   21 The import org.knime.core.data.RowKey is never used   MarvinSpace/src/chemaxon/marvinspace   MarvinSpaceNodeModel.java   line 12   1172161945500   17 The import org.knime.core.node.BufferedDataContainer is never used   MarvinSpace/src/chemaxon/marvinspace   MarvinSpaceNodeModel.java   line 18   1172161945500   22 The import org.knime.core.node.defaultnodedialog.DialogComponentNumber is never used   MarvinSpace/src/chemaxon/marvinspace   MarvinSpaceNodeDialog.java   line 4   1172161945515   23

as well as in the MANIFEST.MF

Code:

Severity and Description   Path   Resource   Location   Creation Time   Id Eclipse-AutoStart header is deprecated, use Eclipse-LazyStart   MolImporter/META-INF   MANIFEST.MF   line 15   1172161946687   25

Nothing serious, I guess, but trying to open the MolImporter/Knime_Project/workflow.knime, even in a KNIME perspective did nothing...


In the Plug-ins view though, I could see the chemaxon and chemaxon.marvinspace present, opening them gave me a webpage-style like with General Info, etc... in the Testing field, I could see a "Launch an Eclipse Application", running it gave me a KNIME perspective and the 2 nodes MarvinSpace and MolImporter in the Node Repository, I could create a KNIME_project (importing from your MolImporter/Knime_Project did not work), add MarvinSpace but the MolImporter node gave me this error:

Code:

Problems occurred when invoking code from plug-in: "org.eclipse.jface". java.lang.NoClassDefFoundError: chemaxon/formats/MolInputStream

(I spare you the rest of the stack trace...)


Now I am stuck - I am probably doing something wrong...


Anyway, your example is a good start for me to see how everything is arranged.


Cheers,


Nicolas

Actually, I could import the workflow you wrote but got this error:

Code:

ERROR    WorkflowManager    Could not find factory class chemaxon.MolImporterNodeFactory for node node_1 ERROR    WorkflowManager    chemaxon.MolImporterNodeFactory ERROR    WorkflowManager    Could not find factory class chemaxon.marvinspace.MarvinSpaceNodeFactory for node node_2 ERROR    WorkflowManager    chemaxon.marvinspace.MarvinSpaceNodeFactory ERROR    WorkflowManager    Could not create connection: connection_1 reason: Source node does not exist in workflow

Launching the Application should be the good way.





I'm trying to figure out what the problem could be here

Code:

Problems occurred when invoking code from plug-in: "org.eclipse.jface". java.lang.NoClassDefFoundError: chemaxon/formats/MolInputStream

Seems like the MarvinBeans.jar is not included in the classpath.


Could you please try Properties from the Package Explorer -> Java Build Path -> Libraries -> Add External Jars, and select MarvinBeans.jar.





I can't see any serious in the warnings, the code just needs a clean-up.

Are the KNIME plugins available for general evaluation? If so how can one download them?

treweren wrote:

Are the KNIME plugins available for general evaluation? If so how can one download them?

Infocom Corporation is preparing the Extensions for Knime by the cooperation of ChemAxon. Please contact us by e-mail.


s.ozawa _at_ infocom.co.jp

There is now a project started on source forge for building and sharing KNIME nodes.


We are starting work on Chemaxon based nodes for KNIME which will be made available on this site.





http://nodes4knime.wiki.sourceforge.net/





The project is currently in planning but will start shortly.





This project is not just for Chemaxon based nodes so please feel free to contribute any nodes into this project.





Regards


Andrew


http://www.edgesoftwareconsultancy.com/

INFOCOM releases ChemAxon's JChem nodes on KNIME workbench.








JChem Extensions contains has about 20 novel nodes for the KNIME workflow platform, from basic functions dealing with chemical structure data (Basic module) to special functions using ChemAxon’s software tools such as Marvin, JChem, Standardizer and more.





The evaluation version is available now.


Please refer to the following pages.





[JChem Extensions]


http://www.infocom.co.jp/bio/develop/jchemextension_en.html





[Academic package: free]


http://www.infocom.co.jp/bio/develop/jchemextension_academic_en.html

Please see http://www.edgesoftwareconsultancy.com/new_nodes4knime.htm





For details of these free to use nodes based on chemaxon components.