Dear CDK-KNIME lovers,
CDK is a mature cheminformatics kit that includes a comprehensive list of molecular descriptors (full list: http://wiki.qspr-thesaurus.eu/w/CDK). Currently “Molecular Properties”, the CDK-KNIME node for descriptor calculation, does not include all of those. Several interesting descriptors already present in CDK project are not currently implemented in KNIME (e.g. CPSA, Kier & Hall descriptors, GravitationalIndex, MDE, MomentOfInertia, WHIM, etc.).
I would like to know if there is a reason why these descriptors were excluded from the node or if they could be implemented in future versions.
I know that CDK-KNIME node already implement what are probably considered the most important but as data miners we're extremely greedy and eager of descriptors. So I think it would be really great to have all of them implemented in KNIME!
Thanks in advance for any feedback.
there is no particular reason other than the one you mentioned why any other descriptors have not been included in the molecular properties node.
I would have to look into the missing descriptors and check their input/output parameters and add them if they appear to be working correctly. However, I don't know how thoroughly some of them have been tested.
It might also be worth to subdivide the implemented set of descriptors into classes.
Do you happen to have an independent test set for all/some of the descriptors?
Thanks for your fast reply. I didn't know that some of the still not implemented descriptors could be not thoroughly tested. If that is the case I would clearly prefer to leave them out from the CDK-KNIME node.
Unfortunately I don't have any independent test set for those descriptors. I don't know neither other independent softwares for their calculation. If someone in the forum know such softwares I propose myself to perform a comparative study for descriptors of those softwares and the CDK ones.