[Rdkit-discuss] New conformer generator: ETKDG

[Sereina <ser...@gm...>]

Dear RDKitters,

Since the release 2015.09.1, a new conformer generator method is available in the RDKit, termed ETKDG. 
The paper describing the method and its performance is published since last week: http://pubs.acs.org/doi/abs/10.1021/acs.jcim.5b00654

To use the method, two flags have to be set when calling the embedding function:
AllChem.EmbedMolecule(mol, useExpTorsionAnglePrefs=True, useBasicKnowledge=True)
or
AllChem.EmbedMultipleConfs(mol, useExpTorsionAnglePrefs=True, useBasicKnowledge=True)

The method uses experimental torsional-angle preferences for a set of SMARTS patterns. If you would like to know which patterns matched your molecule, use the flag printExpTorsionAngles=True.

The experimental torsional-angle preferences (flag useExpTorsionAnglePrefs) and the “basic knowledge”-terms (flag useBasicKnowledge, for a description, see the paper) can also be used separately — this corresponds to the ETDG and KDG methods — but we found the combination (ETKDG) to perform best.

The new methods is slower than standard distance geometry, but the nice thing is that the generated conformers can now be used directly (e.g. the aromatic rings are flat), i.e. no force field minimization is required. Therefore, overall the new method is faster.

I hope you find the new conformer generator useful. If you encounter problems, please let Greg and me know.

Best,
Sereina