from rdkit import Chem
from chembl_structure_pipeline import standardizer
# a molecule with enhanced stereochemistry
m3bis = Chem.MolFromSmiles('Cl.O[C@@H]1CC[C@H](Cl)C[C@@H]1O |a:5,&1:2,8|')
m3bis
(the carbon bearing Cl has absolute configuration, the two carbons bearing OH are part of an 'and' stereo group, so they are present as depicted above + both inverted; so this represents a sample of two diastereoisomers, HCl salts).
len(m3bis.GetStereoGroups())
# 2
m3bis_mb = Chem.MolToMolBlock(m3bis)
m3bis_mb_std = standardizer.standardize_molblock(m3bis_mb)
m3bis_mb_std_mol = Chem.MolFromMolBlock(m3bis_mb_std)
len(m3bis_mb_std_mol.GetStereoGroups())
# 0 # all stereogroups are gone
There might be a more fundamental/theoretical point regarding why one would want to represent mixtures of stereoisomers vs individual stereoisomers.
However, in actual chemoinformatics practice, commercial and corporate samples, including many drugs, often are mixtures of stereoisomers, and they must be handled somehow.
Vendors and companies have long adopted appropriate logic to represent their samples (more) accurately: the chirality flag in V2000 CTAB's to discriminate between ABS and RAC samples; enhanced stereochemistry in V3000 CTAB's for more complex cases like the diastereomeric mixture in the above example; more recently even ChemAxon CXSMILES.
So my question is: is there any rationale behind the removal of stereo groups by standardize_molblock ?
Or is it perhaps related to this similar issue: https://github.com/rdkit/rdkit/issues/6099 ?
And do you know how this can be circumvented ?
There might be a more fundamental/theoretical point regarding why one would want to represent mixtures of stereoisomers vs individual stereoisomers. However, in actual chemoinformatics practice, commercial and corporate samples, including many drugs, often are mixtures of stereoisomers, and they must be handled somehow. Vendors and companies have long adopted appropriate logic to represent their samples (more) accurately: the chirality flag in V2000 CTAB's to discriminate between ABS and RAC samples; enhanced stereochemistry in V3000 CTAB's for more complex cases like the diastereomeric mixture in the above example; more recently even ChemAxon CXSMILES.
So my question is: is there any rationale behind the removal of stereo groups by
standardize_molblock? Or is it perhaps related to this similar issue: https://github.com/rdkit/rdkit/issues/6099 ? And do you know how this can be circumvented ?