Does SMI-TED model support SMILES with stereochemical information?

[dongfeicui]

Hi team,

I noticed that in the function normalize_smiles (at load.py#L41 ), the parameter isomericis set to False by default:

def normalize_smiles(smi, canonical=True, isomeric=False):
    try:
        normalized = Chem.MolToSmiles(
            Chem.MolFromSmiles(smi),
            canonical=canonical,
            isomericSmiles=isomeric
        )
    except:
        normalized = None
    return normalized

As far as I understand, this means stereochemical information in SMILES (such as @, /, or ) will be lost during normalization.

I'd like to confirm:

1. Is the current SMI-TED model designed not to support SMILES with stereochemical information?

2. If that's the case, is there any plan or recommended way to handle isomeric SMILES (eg, enabling isomeric=Trueduring preprocessing)?

Thank you for your time and clarification!

[vichShir]

Hi @dongfeicui ! Thank you for your interest in our work.

I believe your questions is also related to issue #33. We provided some instructions to SMI-TED support SMILES with stereochemical information.

Feel free to reach out if you have any further questions.