KNIME workflow to create RDKit synthon space from SMILES building blocks?

Groups Cheminformatics
,
1

Title: KNIME workflow to create RDKit synthon space from SMILES building blocks?

Hi all, I am reposting this as my earlier post was deleted by me by mistake. I’m looking for a KNIME workflow that takes a set of building blocks (SMILES) and produces an RDKit synthon space file (compatible with rdSynthonSpaceSearch; i.e., the synthon-space text format with synthon SMILES, set/index, reaction id). If you have a sample workflow or pointers (which RDKit KNIME nodes to use), please share. References: https://greglandrum.github.io/rdkit-blog/posts/2024-12-03-introducing-synthon-search.html . Also open to tips using the RDKit nodes for KNIME. Thanks!

2

Hi @Knimeraja -

This one’s not in my wheelhouse. I searched a bit on the KNIME Community Hub and didn’t find anything useful, but maybe that’s just because chemistry’s not my area.

That said, maybe we could tag @greglandrum and see what he might suggest?

4

Thanks, @ScottF, for your reply and for tagging @greglandrum in this post. Recently, I attended one of the KNIME training sessions, and the trainer promised me that they have a workflow for synthon creation and will share it after the training

5

@Knimeraja I don’t have an example laying around, but I’d be happy to work with you to come up with one.

What we need to do this is some sample lists (SMILES or SDF) with the building blocks you want to use and some idea about what the reaction transformation will be.

For example, for a simple amide bond formation from a set of amines and a set of carboxylic acids, that would be one file with the amines to use, one file with the acids to use, and the description would be “the amine N should be coupled to the carboxylic carbon, the carboxylic OH is lost”,

I’m going to be out of the office for a while, but if you can me that information, I will hopefully be able to put something together early next week. (The RDKit UGM is next week, so it may actually end up taking longer than that)

1 Like
6

Thank you very much @greglandrum for your response, and apologies for the late reply. I will prepare the necessary files as per your request and share them with you soon.

1 Like
7

Dear @greglandrum .

Following your request in the KNIME forum, I’ve prepared sample building block files and reaction description for creating an RDKit synthon space workflow.

Files Included:

  1. aryl_halides_building_blocks.csv (1 compounds)

    • Aryl halides (Ar-Cl, Ar-Br, Ar-I) for Buchwald Hartwig coupling
    • Columns: SMILES, ID, Name

  2. amines_building_blocks.csv (25 compounds)

    • Primary and secondary amines for coupling
    • Columns: SMILES, ID, Name

  3. buchwald_hartwig_reaction_description.txt

    • Detailed reaction description and transformation rules
    • SMIRKS patterns and synthon generation guidelines

Reaction Details:

  • Reaction Type: Buchwald Hartwig C-N coupling
  • Transformation: Aryl halide + Amine → Aryl amine + HX
  • SMIRKS: [c:1][Cl,Br,I:2].[N:3][H:4]>>[c:1][N:3]
  • Synthon Generation:
    • Aryl halides: Replace X with [*:1]
    • Amines: Replace N-H with [*:1]

Data Source:

These building blocks are extracted from a dataset of 5,000 validated reaction examples, ensuring chemical validity and diversity.

Expected Synthon Space Output:

The workflow should generate a synthon space file compatible with rdSynthonSpaceSearch with entries like:

# Reaction: Buchwald_Hartwig
[*:1]c1cccc(F)c1    0    BH_001    ArylHalide_001
[*:1]c1ccc(Cl)cc1   0    BH_001    ArylHalide_002
C[N:1]([*:1])c1ccccc1    1    BH_001    Amine_001
[*:1][N:1]Cc1ccccc1      1    BH_001    Amine_002

This provides exactly what you mentioned: “sample lists with building blocks and description of reaction transformation” for amide bond formation equivalent (C-N coupling in this case).

Looking forward to the KNIME workflow development!
Regards,
Nataraj
aryl_halides_building_blocks.csv (79 Bytes)
amines_building_blocks.csv (897 Bytes)

8

Hi @greglandrum , did you get a chance to see my message , thanks !

9

HI @greglandrum , can you help us here ?

10

Hi @greglandrum, apologies for reaching out again, but I truly appreciate your earlier responses and insights. We’re still trying to resolve this issue and would be grateful if you could share any guidance on possible solutions.
If direct help isn’t possible, could you please suggest whether an AI or GenAI-based approach might help address this problem?
Your expertise would mean a lot. Thanks once again for your time and support!