Find zwitterions

[ldamore]

Hello!
Is there any way to find zwitterions in a set of molecules?

[greglandrum]

Sure! If you want to match molecules which have at least one positively charged atom and at least one negatively charged atom, you can use the SMARTS [+].[-]:

In [2]: qry = Chem.MolFromSmarts('[+].[-]')

In [3]: Chem.MolFromSmiles('[NH3+]CC(=O)[O-]').GetSubstructMatch(qry)
Out[3]: (0, 4)

In [4]: Chem.MolFromSmiles('[NH3+]CC(=O)').GetSubstructMatch(qry)
Out[4]: ()

Note that this also matches the charge separated form of nitro groups which the RDKit uses:

In [5]: Chem.MolFromSmiles('CC[N+](=O)[O-]').GetSubstructMatch(qry)
Out[5]: (2, 4)

It only matches atoms which have a +1 or -1 charge... +2 and -2 are rarer, but you could also match higher charge states if you want by using a modfied form of the SMARTS:

In [6]: Chem.MolFromSmiles('[NH3+]CC(=O)[N-2]').GetSubstructMatch(qry)
Out[6]: ()

In [7]: qry2 = Chem.MolFromSmarts('[+{1-}].[-{1-}]')

In [8]: Chem.MolFromSmiles('[NH3+]CC(=O)[N-2]').GetSubstructMatch(qry2)
Out[8]: (0, 4)

[dehaenw]

hi all,
for a very rigid zwitterion filter there are also a few more things that i think can be useful to remove false zwitterion hits

1. a check for 0 net charge (eg usign GetFormalCharge()), because some define zwitterions as net 0 charge species only
2. as greg remarked, opposing charges within a single functional group like in a nitro group are normally not considered zwitterions. this can be addressed by filtering for specific functional groups not considering the charges of some atoms in nitro,azide,isocyanide,amine oxide groups, maybe a few more i forgot about. however, consider that nitronates can form zwitterions e.g. [NH3+]CC=[N+]([O-])[O-] (edit:corrected structure) despite also having an internal + and - charge
3. multicomponent mols need to be split into components (e.g. using GetMolFrags()) because otherwise salts like [NH3+]C.C(=O)[O-] will show up as zwitterions

[ldamore]

Thanks both for your answer, the solution proposed by @greglandrum works great and all the considerations taken helped me to handle zwitterions correctly!

[piermichele-kobauri]

Hi all,
To address point 2 made by @dehaenw, could we match molecules with at least one positively charged atom and at least one negatively charged atom, but separated by at least one bond?
This would result in the SMARTS [$([+{1-}].[-{1-}]);!$([+{1-}]~[-{1-}])]