Hi all, I’m working with several chemical structure sources and it seems like there are some differences between implementations of stereochemistry specification of bridged bicyclic compounds which leads to unlikely structures. For example, one such structure is:
CC(C)(C)OC(=O)N1[C@]2([H])C[C@@](CC2)([H])[C@H]1C3Nc4c(ccc(c4)B5OC(C)(C)C(C)(C)O5)N=3 When I create a molecule from this string with RDKit, I get a rendering which indicates the structure is incorrectly specified (by opposite wedge directions for the hydrogens on the bridgehead carbons). The molecule additionally fails to produce a 3D conformer by `rdDistGeom.EmbedMolecule` (after AddHs). The fix seems to be to correct the original SMILES string by changing one of the bridgehead stereochemistry configurations, which then leads to same-wedge renderings and successful 3D conformer generation. My question is: where do these SMILES strings with “problematic” stereochemistry specifications originate? Are there software implementations of SMILES generation that are internally consistent but incompatible with RDKit’s internal consistency? Does such a disagreement in details originate from ambiguity in the SMILES specification? Best regards, Steve Brown
_______________________________________________ Rdkit-discuss mailing list Rdkit-discuss@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/rdkit-discuss
