Column conversions to rdkit


I am finding the new way rdkit nodes handle chemistry columns somewhat confusing.

the fact that non Rdkit columns can be used as input is great but the node output is most confusing.

For example, I have an SDF column, I connect a Salt Stripper node. In the output I now have the salt stripped molecules in Rdkit format as expected, but now my original SDF column is gone and now converted to Rdkit also, surely this is not expected.

I find myself going back to using all the molecule converter nodes, so that my SDF column can remain untouched.



Hi Simon,

I tend to agree - I would like it if the default action was to append a new column when rdkit automatically accepts non/rdkit input, as I also find myself having to regenerate SDF columns once i am done with rdkit manipulations.


Hi Simon, hi Alastair,

the column that replaces the original SDF column is not the classical RDKit Cell column, but a so-called adapter column. The difference is that the SDF that another node will use from such a column is still the original SDF from your SDF column and not the one that the RDKit cell possesses. So basically, what you are seeing is like a wrapper of other formats that forms around the original column. Why is this disturbing to you? Have you run into cases where the SDF was different from before? If so, that would be a bug and I would be interested in the circumstances. Or is there some other weird behavior of such a column I am not aware of? Please share...



Hi Manuel,

The thing I dont like, is that when its been converted from SDF to RDKit, or as you describe it, wrapped in an RDKit wrapper, the orientation of the molecule has now changed. The visualisation from an SDF to this RDKit wrapped molecule is different with some cpds rotated 180C degrees or flipped left to right. Is it possible for this to be retained correctly?


Hi Simon,

could you please attach me a few sample molecules where this happens with the 180° rotation or flipping? I need to understand if this is a bug (RDKit Nodes would change the SDF) or if it is "just" a rendering thing (KNIME renderer is set to something else than to the normal SDF renderer, e.g. the RDKit renderer which would render correctly the RDKit molecule and not the original SDF anymore) ... Please send me something, and I can think about it with one of the KNIME folks.

Thanks for bringing this up! I hope this can be improved somehow, but my gut feeling is that this would be something we need to workout with the KNIME folks.


It's quite likely the the renderer used is the RDKit renderer and not an SDF renderer. The default renderer will be selected according to the preferred type of the output column and this should be RDKit. And as Manuel said, the RDKit renderer will render the converted molecule.

Looking into this some more, I do see that the renderer becomes the RDKit renderer when it has the RDKit wrapper on the molecule. If I right click on it and force it to use Marvin as the default renderer, suddenly the molecule rotates back 180 degrees to its original position.

So in conclusion, I dont believe the SDF cell is being altered thankfully. Also when I use an RDKit converter node to go back to the SDF celltype, the molecules rotate back to the original positions.

So... is it possible to make the RDKit renderer to draw the SDF format molecules correctly. It is somewhat a pain if I and others will have to keep going into the column header to rechoose Marvin etc to get the desired molecule drawn correctly. It seems as though the RDKit renderer isnt picking up the molecule coordinates correctly.


This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.