[Rdkit-discuss] Replace matched patterns with wildcards

[Marawan H. <mar...@ya...>]

Hi,
I am trying to use rdkit to replace matched SMARTS patterns in a molecule with a wildcard (*), and return a SMARTS string where the original molecule is an instance of this returned SMARTS string,
I tried the following:########from rdkit import Chem
def generate_modified_smarts(smiles, smarts_patterns, num_patterns_to_replace):    molecule = Chem.MolFromSmiles(smiles)    patterns_replaced = 0
    for smarts in smarts_patterns:        if patterns_replaced >= num_patterns_to_replace:            break
        pattern = Chem.MolFromSmarts(smarts)        while molecule.HasSubstructMatch(pattern) and patterns_replaced < num_patterns_to_replace:            match_indices = molecule.GetSubstructMatch(pattern)
            # Extract segments before and after the match            before_match, after_match = "", ""            if match_indices[0] > 0:                before_match = Chem.MolFragmentToSmarts(molecule, atomsToUse=list(range(match_indices[0])))            if match_indices[-1] < molecule.GetNumAtoms() - 1:                after_match = Chem.MolFragmentToSmarts(molecule, atomsToUse=list(range(match_indices[-1] + 1, molecule.GetNumAtoms())))
            # Combine parts with a wildcard            modified_smarts = before_match + '*' + after_match            molecule = Chem.MolFromSmarts(modified_smarts)            patterns_replaced += 1
    return Chem.MolToSmarts(molecule)

example_smiles = "CCOC1=C(C=C2C(=C1)N=CC(=C2NC3=CC(=C(C=C3)OCC4=CC=CC=N4)Cl)C#N)NC(=O)C=CCN(C)C"smarts_patterns = ["C=O", "C#N"]num_patterns_to_replace = 2
modified_smarts = generate_modified_smarts(example_smiles, smarts_patterns, num_patterns_to_replace)print(f"Modified molecule SMARTS pattern: {modified_smarts}")#######
While it seems to work for C=O, it does not for C#N and the connectivity is messed up for C#N, even if I use it alone, i.e. without the carbonyl. The matched patterns could be anywhere in the molecule and could be more complex than this, but I just tried some simple cases to see how robust is this approach. It worked for "CCO", but did not work when i tried "Cl".
I am wondering if this is something you can help with,

Marawan