Re: [Rdkit-discuss] fingerprint a molecule with pseudoatoms denoted by 'Du'

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

Hi Jenke,


I have put together a small gist showing a slightly hacky way to 
round-trip a molecule containing dummy atoms through a PDB block 
(assuming that your molecules do not contain astatine). If your dummy 
atoms are called "DU" rather than " *", you may just change the 
replace() expression with something that fits your needs.


HTH, cheers

p.


On 10/30/19 12:06, SCHEEN Jenke wrote:
> Hi RDKitters,
>
> I'm trying to use rdkit to generate molecular fingerprints (such as AP 
> or ECFP) on molecules that have non-interactive pseudoatoms ('dummy 
> atoms', denoted by Du). I attached a sample PDB file containing the 
> dummy atoms on positions 21-24. Reading this file 
> (Chem.rdmolfiles.MolFromPDBFile("test.pdb", sanitize=False) throws a 
> post-condition violation because the element 'Du' isn't recognised, 
> which makes sense. I've been searching online and haven't been able to 
> find any workarounds, do you have any suggestions?
>
> Some notes:
>
>   * I'm hoping that once rdkit is able to read in the pdb file the mol
>     object can be parsed without the FP constructor (e.g.
>     AllChem.GetMorganFingerprint) complaining.
>   * The use of the term dummy atoms here should not be confused with
>     the dummy atoms depiction in fragmentising molecules in rdkit
>     (where * is the smiles notation).
>   * For this project all I aim to do is generate structural
>     fingerprints for these types of ligands. This means I won't have
>     to worry about defining chemical properties to Du.
>   * The context for this issue is that we're aiming to featurise the
>     ligands for an ML protocol where the dummy atoms are one of the
>     major descriptors of the problem.
>
>   * I thought manually inserting a 119th element in atomic_data.cpp
>     might resolve the issue but I've been unable to locate the file in
>     my conda installation.
>   * The ODDT python API seems to parse the Du element without any
>     issues but is limited in its FP generator diversity.
>
>
> Best,
>
> Jenke
>
> The University of Edinburgh is a charitable body, registered in 
> Scotland, with registration number SC005336.
>
>
>
>
> _______________________________________________
> Rdkit-discuss mailing list
> Rdk...@li...
> https://lists.sourceforge.net/lists/listinfo/rdkit-discuss