I was having trouble with 5-membered heterocycles failing to sanitize using the RDKit chemical transformation node. A workaround was to change the aromatic smiles bonds to alternating double and single bonds. The file reader in versions prior to 4.3.0 used to take these modified smiles and convert them to molecule objects having the Kekule-type bonds, which process normally in RDKit’s transformation node. With the latest KNIME version file reader it transforms the fixed smiles into having aryl bonds. I don’t seem to have a way to fix this anymore. Anyone else having the problems with 5-membered heterocycles e.g. triazoles.
Hi,
I’m a bit confused by this because the file reader doesn’t actually know any chemistry and doesn’t parse the SMILES.
Could you please share a sample workflow that demonstrates the problem?
Thanks,
-greg
Hi Greg, It turns out that I had to change my file paths for all of my workflows and I had incorrectly pointed to a file in one node. Everything is as before. As you know the file reader can specifically read a text string as smiles if designated. So it has some sort of interpreter, perhaps wrapped into RDKit or something else. Thanks, Brad
Hi Brad,
As you know the file reader can specifically read a text string as smiles if designated. So it has some sort of interpreter, perhaps wrapped into RDKit or something else.
This isn’t quite true, but it’s easy to understand how the misconception arises.
When you change the type of that column in the file reader dialog from “string” to “SMILES”, KNIME just changes the type of the column. It continues to read that column from the file using the standard text parsing and stores exactly the same data in the table. The result is exactly the same as if you left the column as a string column and later used the molecule type cast node to change the type of the column (that one also just changes the type, it doesn’t do any chemistry).
It’s easy to get confused because if you have some of the cheminformatics extensions installed you see pictures of molecules when you look at tables containing a SMILES column. This is because some of the cheminformatics extensions install a “renderer” which tells the Analytics Platform that it can generate pictures for smiles columns. If you want to see the original SMILES you can turn the renderer off by right-clicking on the column header when viewing the table and selecting “String” from the dialog that appears.
This seems like it’s just a detail, but it is frequently helpful to know that KNIME doesn’t do any chemistry perception or modify the data in SMILES, SDF, or other molecule cells in any way.
-greg