Interpreting the results of controllingAtoms (stereoAtoms) in StereoInfo

[najwalb]

I have trouble interpreting the results of the controllingAtoms vector as returned by FindPotentialStereo() in the StereoInfo object. The stereoatoms identified by rdkit automatically from smiles strings make are easier to interpret, but I need to use FindPotentialStereo() in my application since I am building molecules using RWMol() and need to add the stereoinformation post-hoc.

Here is an example to illustrate:

cano_smi = '' # put smiles from the options given below
m = Chem.MolFromSmiles(cano_smi)

stereo_info_double_bond = [e for e in Chem.rdmolops.FindPotentialStereo(m) if e.type==Chem.rdchem.StereoType.Bond_Double]
stereo_description_double_bond = {e.centeredOn:e.descriptor for e in stereo_info_double_bond}
print(f'stereo_description_double_bond {stereo_description_double_bond}\n')
stereo_atoms_from_cip = {e.centeredOn:[c for c in e.controllingAtoms] for e in stereo_info_double_bond}
print(f'stereo_atoms_from_cip {stereo_atoms_from_cip}\n')

bond_stereo_from_smi = {bond.GetIdx():bond.GetStereo() for bond in m.GetBonds() if bond.GetStereo()!=Chem.rdchem.BondStereo.STEREONONE}
print(f'bond_stereo_from_smi {bond_stereo_from_smi}\n')
stereo_atoms_from_smi = {bond.GetIdx():[a for a in bond.GetStereoAtoms()] for bond in m.GetBonds() if bond.GetStereo()!=Chem.rdchem.BondStereo.STEREONONE}
print(f'stereo_atoms_from_smi {stereo_atoms_from_smi}\n')

• cano_smi = 'CCOC(=O)C/C(=C\c1ccc(N)c(C)c1[N+](=O)[O-])C(=O)OCC'

stereo_description_double_bond {6: rdkit.Chem.rdchem.StereoDescriptor.Bond_Cis}

stereo_atoms_from_cip {6: [5, 19, 8, 4294967295]}

bond_stereo_from_smi {6: rdkit.Chem.rdchem.BondStereo.STEREOE}

stereo_atoms_from_smi {6: [19, 8]}

• cano_smi = 'CCOC(=O)/C=C/c1cc(CN(C)C)ccc1Cl'

stereo_description_double_bond {5: rdkit.Chem.rdchem.StereoDescriptor.Bond_Trans}

stereo_atoms_from_cip {5: [3, 4294967295, 7, 4294967295]}

bond_stereo_from_smi {5: rdkit.Chem.rdchem.BondStereo.STEREOE}

stereo_atoms_from_smi {5: [3, 7]}

• cano_smi = 'C\C=C(/C)C=O'

stereo_description_double_bond {1: rdkit.Chem.rdchem.StereoDescriptor.Bond_Cis}

stereo_atoms_from_cip {1: [0, 4294967295, 3, 4]}

bond_stereo_from_smi {1: rdkit.Chem.rdchem.BondStereo.STEREOE}

stereo_atoms_from_smi {1: [0, 4]}

Note that each time the 'correct' (i.e. matching the smiles' identification) atom indices are identified by FindPotentialStereo() but they are given in different locations in the controllingAtoms vector alongside the placeholder value 4294967295. I'd like to know: 1) how to interpret the controllingAtoms vector in general, and 2) how to retrieve the atom indices matching the ones recovered by rdkit from the smiles string.

[greglandrum]

@najwalb The controlling atoms in a StereoInfo object (the objects returned by FindPotentialStereo()) are in the order:
[begin_atom_neighbor1, begin_atom_neighbor2, end_atom_neighbor1, end_atom_neighbor2]
If either begin_atom or end_atom have only one neighbor then their neighbor2 slot can have the placeholder value Chem.StereoInfo.NOATOM (which is 4294967295).
I don't think there should be any cases where either begin_atom_neighbor1 or end_atom_neighbor1 are equal to Chem.StereoInfo.NOATOM

To interpret the values:

• Bond_Cis means that begin_atom_neighbor1 and end_atom_neighbor1 are cis with respect to each other (on the same side of the double bond)
• Bond_Trans means that begin_atom_neighbor1 and end_atom_neighbor1 are trans with respect to each other (on opposite sides of the double bond)