Re: [Rdkit-discuss] Washing molecules in RDKit

[Greg L. <gre...@gm...>]

Dear JP,

On Thu, May 12, 2011 at 12:31 PM, JP <jea...@in...> wrote:
>
> I was wondering if there was a way to wash molecules in RDKit?

Anything is possible given sufficient code. ;-)

Jokes aside, as it stands there's not much RDKit functionality for
automatically cleaning up molecules. Details below.

> By washing, the manual in MOE means (and I understand that your definition
> might vary to some degree):
>
> Replace or recalculate the molecule name

You can replace the name easily by doing :
  mol.SetProp("_Name","new name")
but there's no capability to automatically generate an IUPAC name or
anything like that

>
> Disconnect simple metal salts drawn in covalent notation

This would be pretty easy to do using either a reaction or an EditableMol.

> Remove minor components (e.g. counterions and solvent molecules), optionally
> storing the removed components in an alternate field

Salt removal, which could also do solvent removal, is in
rdkit.Chem.SaltRemover. The code uses a library (defined by SMARTS) of
things to remove which is read from a text file (by default
$RDBASE/Data/Salts.txt). There's currently no ability to add the
removed pieces to an alternate field, but one could code that.

It's simpler, though less accurate, to just split the molecule into
fragments using Chem.GetMolFrags and then just keep the largest
fragment.

>
> Rebalance protonation states by deprotonating strong acids and/or
> protonating strong bases

This would be easy to do given a set of SMARTS patterns defining the
strong acids/bases.

> Add explicit hydrogen atoms, or remove safely-deletable explicit hydrogen
> atoms

Adding and removing Hs is easy, but there's currently no concept of
"safely deleteable". When you call Chem.RemoveHs it takes just about
everything (it won't remove an H from H2 or leave the molecule empty).
If you can define what you mean by safely deletable, it probably would
be easy to add.

>
> Replace the atomic coordinates with an aesthetic 2D depiction layout

yes. AllChem.Compute2DCoords

>
> Enumerate tautomers and possibly also protonation states

no

> Filter enumerated structures according to strong acid/base rules

Again, given a library of strong acid/base definitions this would be easy.

Best,
-greg