CDK-Taverna workflow

We are glad to announce that our article about CDK-Taverna, an open workflow solution for cheminformatics, is now online on BMC Bioinformatics. CDK-Taverna, which lives at http://www.cdk-taverna.de/, features more than 160 workers for various tasks in molecular informatics.

The workflow paradigm allows scientists to flexibly create generic workflows using different kinds of data sources, filters and algorithms, which can later be adapted to changing needs. In order to achieve this, library methods are encapsulated in Lego(TM)-like building blocks which can be manipulated with a mouse or any pointing device in a graphical environment, relieving the scientist from the need to learn a programming language. Building blocks, so-called workers, are connected by data pipelines to enable data flow between them, which is why pipelining is often used interchangeably for workflow.

Now, this does of course include the chemical structure editor mentioned above. As a pure coincidence , I had started a chemical structure editor in the late 1990s, called JChemPaint, which was later influenced a lot contribution from Egon Willighagen and others. JChemPaint experienced several rounds of re-architecture but for some reason it was never really considered to be production-ready.

We are now taking another go at it, at its applet version to be precise. And this time my charming profession team will give it a try and I’m confident that this time we’ll get it right. The fact that we are not alone in developing JChemPaint is both boon and bane. Everyone in an open source project is happy about harvesting the man-power of others but on the other hand you loose some control. While we, for example, are looking at the applet from a usability and GUI perspective, Egon and Arvid at Uppsala do a re-architecturing of the controler (the piece of software dealing with the user input) in order to make it maintainable through modularization.

So, in order to get an idea of what we actually need, we decided to do an analysis of a) existing structure editor applets and what they to well and not so well and b) how our curators do their work. For the latter, we did a screen recording of our curators’ work and looked at the type of functionality they used and how they did this (keyboard shortcuts, mouse, buttons, menus). The preliminary results are available on the JChemPaint wiki and we are more than happy to receive input from the community about this.

The workshop itself was phantastic – we had 40 participants with well balanced contributions from industry and academia. The first half day was dedicated to tutorials on various aspects of CDK, basic installation, our CDK and Oracle-based chemical search engine, scripting of CDK from within Bioclipse and CDK-based workflows in KNIME. The second day brought talks about science and applications based on CDK by John van Drie, Asad Rahman und Oliver Karch.  I reported earlier on those talks.

This official part was followed by a developers workshop in the format of an unconference.  Participants suggested topics on the fly and then voted with their feet. The topics that then made it where:

• Clojure
• Mining ChEMBL with CDK pharmacophore stuff
• Threading group
• Top 10 improvements (CDK WS 2009 unconference)
• JChemPaint/Rendering

all of which reported their results on the CDK wiki.

Another important outcome was a movement to join forces in creating and maintaining joint CDK-based plugins for KNIME and Bioclipse. There is a lot of synergy to be harvested. Licensing problems will need to be overcome but that seems doable.

A nice thing about molwind is that is they have open-sourced it. It lives at molwind.org

In a second part, he jumps to explaning a typical use case, a pharmacophore-based virtual screening setup. After a historical view back, he showed his work with Rajarhi Guha on CDKPsearch. John and Rajarshi have created an XML pharmacophore description, which allows to define a pharmacophore,  walk through an SDF and find fitting compounds. John has used the package in three different application scenarios with three different companies.

To finish, John brings up the question what we would need to use a pharmacophore setup to mine the data in ChEMBL, the new Chemgenomics Resource at EBI.

For all of you out there, I’ll urge you to download KNIME from http://www.knime.org/ and try it out.

• Bioclipse
• The Use of CDK in Bioclipse
• Scripting in Bioclipse
• Calculating CDK properties

First we’ll download Bioclipse 2.0Beta4 from http://www.bioclipse.net.

Ola continues by explaining Bioclipse’s component architecture. For the tutorial part of his talk, he has prepared an extended set fo phantastic cheet sheats, which the user can simply run and which translate into Bioclipse scripts which are downloaded via a GIST repository. These cheat sheets can be downloaded via Bioclipse’s software update feature and I urge you to try them out.

Mark points out a few performance issues (finding all rings in a Bucky Ball ).

He then goes into how (and why) to use a combination of the CDK fingerprinter and the CDK subgraph ismorphism to achieve a good performance in chemical database searching. Mark has put up an exercise on http://www.ebi.ac.uk/~markr.

Feel free to go ahead and fill the gaps.

Egon has prepared a LaTeX document with his teaching material as well as the code examples on at http://pele.farmbio.uu.se/groovy. Turns out that Groovy scripting is a really nice environment for writing CDK code. You can say things like

import org.openscience.cdk.interfaces.*;
import org.openscience.cdk.*;
import org.openscience.cdk.atomtype.*;
import org.openscience.cdk.config.*;
import org.openscience.cdk.tools.manipulator.*;
import javax.vecmath.Point3d;

molecule = new Molecule();
atom = new Atom(Elements.CARBON);
molecule.addAtom(atom);
matcher = CDKAtomTypeMatcher.getInstance(
DefaultChemObjectBuilder.getInstance()
);
type = matcher.findMatchingAtomType(molecule, atom);
AtomTypeManipulator.configure(atom, type);
println “Atom type: $type.atomTypeName”

As you can see, you do not need to handle things like Exceptions or Data Types. Groovy, like most other scripting environments will handle that for you.

If CDK.jar is in your CLASSPATH, you can run this code inside GroovyConsole and you’ll get “C.sp3″ as an output.

Egon’s course material have many more examples.