Dear Anthony,
Indeed, Au-S or As-O bonds have to be explicitly defined, otherwise how is
the code supposed to know where the metal atom should be?.
Regarding the geometry of organometallics, try generating coordinates in
RDKit but do not optimize with MMFF (or UFF). UFF (in principle) handles
metals but don't expect miracles.
I've just tried building ethylthiogold in KNIME. "RDKit Generate Coords"
(corresponding to EmbedMolecule) builds rather nice initial coordinates,
which might be good enough as a starting point for b97-3c (or, even better,
CREST followed by b97-3c). BTW, MolConverter from the MarvinBeans suite
generates pretty decent geometries of metal complexes.
Best regards,
Michal

On Tue, 1 Dec 2020 at 15:21, Anthony Nash <anthony.n...@ndcn.ox.ac.uk>
wrote:

> Thanks Michal and Jan. Much appreciated.
>
> As per your comments, of the 140 structures I had parsed over, the
> AllChem.MMFFOptimizeMolecule() was returning a -1 (just spotted the
> documentation included a return value - my bad!) on four of those
> structures to suggest "Force field could not be setup". Those four
> contained a combination of Au or As elements. The quick reasonable
> structure calculation was then going to be performed with a B97-3c level of
> chemistry using ORCA.
>
> Unfortunately, prior to the ORCA step, initial placement of some elements
> is too far from where one would expect them to be for an QM calculation to
> have produced a recognisable structure . As Jan suggested, I'm beginning to
> think that explicitly defining the Au-S and As-O bonds might be required!
>
> Thanks
> Anthony
>
> ------------------------------
> *From:* Michal Krompiec <michal.kromp...@gmail.com>
> *Sent:* 01 December 2020 15:07
> *To:* Anthony Nash <anthony.n...@ndcn.ox.ac.uk>
> *Cc:* RDKit <rdkit-discuss@lists.sourceforge.net>
> *Subject:* Re: [Rdkit-discuss] Incorrect gold particle placement
>
> Dear Anthony,
> From MMFF or UFF you shouldn’t expect reasonable structures of
> metal-containing compounds. If you need a quick and qualitatively OK
> method, I recommend xtb.
> Best wishes,
> Michal Krompiec, Merck KGaA
>
> On Tue, Dec 1, 2020 at 1:09 PM Anthony Nash <anthony.n...@ndcn.ox.ac.uk>
> wrote:
>
>
> Dear all,
>
> I'm new to RDKit and cheminformatics in general. I'm using the latest
> RDKit libraries. Any suggestions you can offer would be kindly received.
>
> Using the canonical SMILES   C(C1C(C(C(C(O1)[S-])O)O)O)O.[Au+]   of
> Aurothioglucose  (Pubchem CID: 454937) I've generated a 3D structure
> using the python RDKit code:
>
> mol = Chem.MolFromSmiles(self.canonicalSMILES)
> mol3D = Chem.AddHs(mol)
> AllChem.EmbedMolecule(mol3D)
> AllChem.MMFFOptimizeMolecule(mol3D)
>
> Unfortunately, the mol3D representation has Au+ right in the middle of and
> in the plane of the benzene rings, too far from the negatively charged
> sulfur. I'm new at generating structures from SMILES. Are there any steps
> I'm missing that could help correct the placement of Au+?
>
> Thanks
> Anthony
>
> _______________________________________________
> Rdkit-discuss mailing list
> Rdkit-discuss@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/rdkit-discuss
>
>
_______________________________________________
Rdkit-discuss mailing list
Rdkit-discuss@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/rdkit-discuss

Reply via email to