latest `rdkit` version is not compatible with `chebai` / SMILES canonicalisation needs to be updated

[sfluegel05]

I had some problems running Chebifier with the latest rdkit version (2024.9.6):

ImportError: cannot import name 'MolToMPL' from 'rdkit.Chem.Draw' (.venv2\Lib\site-packages\rdkit\Chem\Draw\__init__.py)

MolToMPL is used in result/molplot.py to create attention plots.

I was able to fix this by reverting rdkit to version 2024.3.6

Todo

• Check if there is an alternative to MolToMPL (maybe rdkit just renamed it?)
• If not, put rdkit <=2024.3.6 in the requirements

[aditya0by0]

rdkit/rdkit#7176
As suggested by above issue, the blog (https://greglandrum.github.io/rdkit-blog/posts/2024-01-18-molecules-on-plots.html) will be helpful in this case.

The below is new alternative to MolToMPL in latest version.

from rdkit.Chem.Draw import rdMolDraw2D

References:

https://www.rdkit.org/docs/cppapi/classRDKit_1_1MolDraw2D.html
https://www.rdkit.org/docs/source/rdkit.Chem.Draw.rdMolDraw2D.html

[sfluegel05]

From #138 (@aditya0by0)

RDKIT behavior changed in latest version

raw_data = CC(=O)NC1CC1[Mg-2]

                mol = Chem.MolFromSmiles(raw_data.strip())
                if mol is not None:
                    raw_data = Chem.MolToSmiles(mol, canonical=True)

rdkit 2024.3.6 returns CC(=O)NC1CC1[Mg-2] (version on my local machine) rdkit 2025.9.3 returns CC(=O)NC1C[CH]1[Mg-2] (latest version)

This is the reason causing failure of the test.

@sfluegel05 Do we need to pin the rdkit version (as per #83) or should we look for an upgrade ?

Short term: I would agree with pinning rdkit to 2024.3.6. Long term: We should fix the drawing function and SMILES canonicalisation (probably using libRDChEBI

[sfluegel05]

see libRDChEBI/formats.py in https://github.com/chembl/libRDChEBI

[aditya0by0]

fixed version of rdkit no compatible with latest python

(python-chebai-graph) sh-4.4$ uv pip install -e .
Using Python 3.14.3 environment at: /home/staff/a/akhedekar/python-chebai-graph/.venv
  × No solution found when resolving dependencies:
  ╰─▶ Because rdkit==2024.3.6 has no wheels with a matching Python ABI tag (e.g., `cp314`) and chebai==1.3.0 depends on rdkit==2024.3.6, we can conclude
      that chebai==1.3.0 cannot be used.
      And because only chebai==1.3.0 is available and you require chebai, we can conclude that your requirements are unsatisfiable.

hint: You require CPython 3.14 (`cp314`), but we only found wheels for `rdkit` (v2024.3.6) with the following Python ABI tags: `cp39`, `cp310`, `cp311`, `cp312`

[sfluegel05]

well, we also don't claim to run on the latest Python version, so this is kind of expected. See #66