I would like to inform you that we updated the RDKit Binaries used in the RDKit Nodes and implemented some smaller bugfixes and performance improvements in the Nightly Build. It is getting time for a new release for KNIME 2.8 and 2.9, and we feel that the current feature set and the bugfixes done since the last release will make a difference to our users. Therefore, I increased the version number to 2.4.0.
If you find the time to test the latest version it would be great to get some feedback. Without serious show stoppers a realistic release date could be still before Feb 12.
Thank you for using and supporting the RDKit Nodes!
Everything seems to be working on the latest RDKit nightly build on KNIME 2.9.1 on the Mac.
I do have one question as an aside for future work, the Gasteiger charges node is really quite a nice addition but find its implementation maybe a bit limited.
It would be quite user friendly if the "RDkit Highlight Atoms" node could support the output from this "Calculate Charges" node and provide an SVG image of the molecule with the Gasteiger charges next to each atom (may be to a couple of decimal places).
A second useful feature of the Gasteiger charges can be in the predictions of pKas of nitrogens. Now I know it wont be perfect, but it will be a start, as I notice some users enquiring about an open source pKa predictor node. If the output of the node somehow allowed you to distinguish which charges are associated with a Nitrogen atom, you could use a Linear Regression model in KNIME against known pKa measurements and these Nitrogen charges to build a model and predict future pKa's.
Also would there be plans to extend the "RDkit Calculate Charges" node to support other atom charge calculations such as Natural NPA, Lowdin, and AIM charges which can be considered to be better pKa predictors.
We'll have to look into adding custom atom labels in the highlight atoms code. It's not trivial to do so, but there's probably something we can do.
You can already find the charges associated with nitrogens using a combination of caleculate charges, substructure filter, and a java snippet. The attached workflow shows how.
We have no immediate plans to add support for quantum-mechanical charges. We could probably add MMFF94 charges, but those would have the liability of only working for molecules where all atoms have MMFF94 parameters.
I have no idea how that blasted Java Snippet works, I cannot get my head around how those Java instructions work, but I can see its simply taking the nth number in the list from the Substructure search results.