Is it possible to reintroduce TPSA in the molecular properties node? and There are plans to introduce the other cdk descriptors?



yes, it is. I have implemented the TPSA descriptor. It should now be available as part of the Molecular Properties node (nightly build).

The descriptors are added on the fly depending on demand. If you wish any other descriptors to be included, feel free to send us a message. :-)



Dear Stephan,

If it is possible i think that will be usefull to have of course included in 1.4 cdk CPSA all,Vabc, BCUT, moments of inertia, autocorrelation moreau-broto and of course alogp and Mlogp.

I don't know if this colud be done for all of them.

thanks a lot  in advance 

Hi Stephan,

Asking for request for new descriptors to be added is risking the floodgates being opened for additions!! :-)

I'll get mine in now;

- Most basic pKa in the molecule

- Most acidic pKa in the molecule.

- cLogD.

- A few different cLogP measures, always worth having a few cLogP descriptors to find one which correlates best to the data.

- sp2 or sp3 character percentage.

- Molecular Formulae.

- Heavy Atom count.

- No of chiral centres.

I hope some of these are possible.



dear stephan,

In my version of knime (update 2.5) the TPSA is still not present.  i run on OS winxp 32. when will be available ?


for simon,

you right these descriptor can help also but I choose this because are included in CDK 1.4 library (look at the site to know which are implemented). Some of them are implemented in indigo node 

Hi Fab, Simon,

I will try to implement some of the descriptors mentioned above over the next couple of days and let you know the outcome.

I should emphasise that I can only implement descriptors that are in the CDK package. The KNIME nodes wrap CDK's functionality.




aplogies for the delay. I have implemented the following descriptors:

  • CPSA
  • TPSA
  • Manhold
  • Autocorrelation
  • Vabc
  • BCUT
  • ALogP
  • Molecular Formula
  • Heavy Atom Count

Some of the above return a list of values. Those are represented in KNIME as ListCell. To expand the ListCell you can use Split Collection Column. 

Any feedback about the implementation and possible glitches would be most welcome.

The updated node is available via the nightly build. @Simon, the TPSA didn't get through last time. It should now be up in the repo for sure.

Best wishes,


dear stephan,

Great job, it is a great improvments for us. The only thing is that starting with the version 2.4.2 updated untill 2.5 when i do the udate for the nightly buld this new implementation is not matched and i have to install the new 2.5 version. Is it possible to fix this ? i have a lot of workflows in the previous one....

thanks agian

wonderfull job


You mean the old CDK integration from the default KNIME Update Site? This one is not maintained any more. Still, you should be able to open workflows with old CDK nodes when you have only the new CDK plug-in from the Community Contributions installed. The old nodes should automatically be replaced by the new nodes. The only thing that cannot be kept is the data due to incompatible serialization formats between the two versions.

dear stephan,

I found that CPSA dosen't work well i used the same file compared with cdkgui calculator and in the knime mode the results is NaN for all but in cdkgui works well! Can you check the descriptor calculation for CPSA? all the other descriptors works well 

Hi Fab,

thanks for the feedfback. CPSA only works on molecules with 3D coordinates. Since the same file works with your cdkgui, may I assume that your sdf file contains 3D coordinates? If so, please send me an example and I'll try to figure out why it does not with the node.

The node works well with the 3D structures I got from ChemSpider.



dear stephan ,

i attached the sdf ..and attached the txt from cdkgui with the results , of course in knime NaN

Dear Fab,

your SDFile does not contain molecules with 3D coordinates. The cdkgui treats your molecules as flat 3D molecules and hence calculates some values for the CPSA descriptor, which can only process 3D molecules.

You can achieve the same result as the cdkgui (almost) if you do the experiment of adding a tiny z coordinate to one atom of each molecule in your SDFile. Something like 0.0001. That way you pretend you have a 3D molecule because the sum total of the z coordinates is not null any more. However, naturally the CPSA result you get back would be wrong.

Please use fully configured molecules with 3D coordinatese for the CPSA descriptor.