[Rdkit-discuss] preservation of stereocenters in 2D->3D generation

[Marshall Levesque]

Greg-

Looping over the atoms of 1000 molecules with  
mol.GetAtomWithIdx(i).GetChiralTag() gave me the statistic of ~58%  
stereocenters being preserved after the molecules were embedded (2D- 
>3D), creating multiple conformers (about 2 conformers per rotatable  
bond) and UFFOptimized.  So my questions are:

A) Does RDKit's embedding process preserve stereochemistry?

B) If not, is the best way to do preserve the stereochemistry by  
looping over the atoms in each 3D conformation produced and checking/ 
comparing to the original 2D structure with GetChiralTag()?

Here's the code for whoever wants to try it out:

#### chiraltest.py   ####

suppl = Chem.SDMolSupplier(infilename)

for j,mol in enumerate(suppl):
  if mol is None:
    continue
  else:
    molname = mol.GetProp(' MOLECULE_ID ')
    for i in range(mol.GetNumAtoms()):
        val = mol.GetAtomWithIdx(i).GetChiralTag()
        if val != Chem.rdchem.ChiralType.CHI_UNSPECIFIED:
            print 'Atom: ',i, 'on ', molname, 'is ', val

##################

Thanks!

-Marshall