# QM9 dataset with chirality

I am trying to find chiral centers of molecules in QM9 dataset. Browsing their SMILES representation, I noticed SMILES yield using

datasets.get_qm9(GGNNPreprocessor(), return_smiles=True)

don't have @ or @@, but it seems that some of these molecules have alternative representations. For instance, both 'Cc1ccoc1C(C)O' and 'CC@Hc1c(C)cco1' represent the same molecule, but of course the latter one outputs the chiral center:

[(1, 'S')]

while only the former is available in the dataset. I found this at https://github.com/chemspacelab/qm9-C7O2H10. Does anyone know if this can be generalized to all the molecules in the QM9 dataset? Thanks!

• I'd be very, very careful using random versions of QM9. The original files are available albeit a bit non-standard - but your problem is based on incorrect stereo-stripped SMILES - however you got them. – Geoff Hutchison Sep 21 at 20:14
• @GeoffHutchison Cc1ccoc1C(C)O is from chainer_chemistry package. CC@Hc1c(C)cco1 is from github account above, which I realized that belongs to one of QM9 authors (von Lilienfeld). – Blade Sep 24 at 15:02

I don't know about that particular GitHub archive, although I'll point out that the smiles.csv contains several kinds of SMILES - the final column containing the correct SMILES.

IMHO, this is an unfortunate result of people mis-using SMILES. As you note, it's perfectly possible to take a chiral molecular structure and output a SMILES with no stereochemical information.

Take alanine - to a chemist, if you simply say 'alanine,' they immediately think 'amino acid' and translate to L-alanine, or C[C@@H](C(=O)O)N as the so-called "isomeric SMILES":

Unfortunately, if you go to PubChem for L-Alanine, you'll see a "canonical SMILES" which sounds official and is CC(C(=O)O)N - no stereochemistry!

The official version of QM9 can be found on FigShare based on the corresponding publication.

While the official files are in a somewhat non-standard XYZ format, they do have the stereochemically-correct SMILES at the end of the file, e.g. C1N2[C@H]3[C@@H]4[C@@H]5O[C@]13[C@H]2[C@H]45 Just watch that some numbers are formatted like -8.0386*^-6 (i.e., $$-8 \times 10^{-6}$$) and can sometimes be misinterpreted by programs like Open Babel or RDKit .. practically everything I've seen.

• Thanks for sharing the original source. Although I was familiar with the publication, I never checked their original dataset for some reason, maybe never thought it would be any different. I'll try see if the results improve with this. – Blade Sep 24 at 15:05