Re: [Rdkit-discuss] Generation of stereo-isomers

[Paolo T. <pao...@un...>]

Dear Soren,

I have recently used the RDKit to enumerate stereocentres.
The approach I followed was to generate a 3D structure for the molecule 
of interest (including hydrogens) using EmbedMolecule(), followed by 
MMFF optimization; do not pay attention to stereochemistry at this 
stage. Then I used FindMolChiralCenters() to find the n chiral centres 
in the 3D structure thus obtained; there are 2^n theoretical possible 
diastereomers. You may enumerate all of them with a "for i in range (0, 
2^n)" loop; converting i to binary, and considering 0 for S and 1 for R, 
you may easily find all possible combinations. Use SetChiralTag() to set 
the chirality on each stereocentre, then rebuild 3D coordinates with 
EmbedMolecule() followed by MMFF optimization. Carry out a final check 
to verify that the chirality of the molecule is the one you initially 
required; if not, try to rebuild 3D coordinates. If you fail, say, 20 
times in a row, it means you are attempting to build a diastereomer that 
cannot give rise to a sensible 3D geometry (such as the case mentioned 
by Peter); then give up.
You will also need to check if any of the structures that you obtain 
during the enumeration has more chiral centres than the initial one (for 
which you didn't specify steeochemistry); in fact, it could be that you 
started from a meso structure. If this happens, just restart enumeration 
from scratch with the increased number of stereocentres.
I also carried out some geometry checks on the generated stereocentres 
coupled with MMFF energy evaluation relative to the most stable 
diastereomer to make sure that the diastereomers obtained where 
reasonable and not unreasonably strained.
This brute-force enumeration is easy to implement, reasonably fast, and 
it worked very well for me.

Kind regards,
Paolo

On 24/09/2015 22:17, Soren Wacker wrote:
> Hi,
>
> is it possible with RDKit to generate all stereoisomers of a given 
> compound?
>
> If not, is anyone working on it?
>
> If not, how difficult would it be / what would be the best way to 
> implement such a function.
>
> best regards
>
> Soren
>
>
> ------------------------------------------------------------------------------
>
>
> _______________________________________________
> Rdkit-discuss mailing list
> Rdk...@li...
> https://lists.sourceforge.net/lists/listinfo/rdkit-discuss