If you're willing to live with the RDKit's definition of bridgehead (see
below), then there is built-in functionality you can use:
from rdkit.Chem import rdqueries
qa = rdqueries.IsBridgeheadQueryAtom()
mol = Chem.MolFromSmiles('C1CC2CCC1C2')
mol.GetAtomsMatchingQuery(qa)
That last call returns a sequence with the matching atoms.
The RDKit bridgehead definition:
// at least three ring bonds, all ring bonds in a ring which shares at
// least two bonds with another ring involving this atom
is definitely not perfect, primarily because of the use of the ring
systems, but it's the best that we were able to come up with while keeping
things efficient.
There's some discussion here https://github.com/rdkit/rdkit/pull/6061 and
in the linked issue.
-greg
On Fri, Aug 25, 2023 at 11:23 PM Wim Dehaen <wimdeh...@gmail.com> wrote:
> greetings all,
> i have thought about the problem some more, and in the end came to the
> conclusion that looping through all rings really is necessary. In the gist
> below you can see the adjusted code, making use of Pat Walters' method
> <https://sourceforge.net/p/rdkit/mailman/message/30387811/> for finding
> all rings. Apologies for the code being messy.
> https://gist.github.com/dehaenw/41eb8e4c39c1158e88b36c6dfc2606d8
> fortunately, this one manages to also detect these difficult cases, see
> below:
> i did not check how fast it is, but i guess it will be a fair bit slower.
>
> best wishes,
> wim
>
> On Fri, Aug 25, 2023 at 8:28 PM Wim Dehaen <wimdeh...@gmail.com> wrote:
>
>> Dear Andreas,
>> that's a good find. i agree the breaking case can be considered
>> bridgehead structure, as it's essentially bicyclo-[3.2.1]-octane plus an
>> extra bond. I need to think about this some more, but it might be related
>> to getting the ringinfo as SSSR instead of exhaustively. The best solution
>> may therefore be to just prune non ring atoms from the graph, enumerate all
>> rings and check it really exhaustively.
>> FWIW: rdMolDescriptors.CalcNumBridgeheadAtoms(mol) returns 0 for mol =
>> Chem.MolFromSmiles("C1CC2C3C2C1C3") too, so this may be an rdkit bug on
>> this end.
>> best wishes
>> wim
>>
>> On Fri, Aug 25, 2023 at 5:20 PM Andreas Luttens <
>> andreas.lutt...@gmail.com> wrote:
>>
>>> Dear Wim,
>>>
>>> Thanks for your reply!
>>>
>>> Apologies for the delay, finally got time to pick up this project again.
>>>
>>> Your suggestion works great, though I have found some cases where it
>>> breaks. For instance the molecule:
>>>
>>> mol = Chem.MolFromSmiles("C1CC2C3C2C1C3")
>>>
>>> It seems, in this case, a bridgehead atom is also a fused-ring atom.
>>> Maybe these looped compounds have too complex topology for this type of
>>> analysis.
>>>
>>> I don't see a straight way forward to identify just the bridgehead atoms.
>>>
>>> Best wishes,
>>> Andreas
>>>
>>> On Sat, Dec 3, 2022 at 12:53 PM Wim Dehaen <wimdeh...@gmail.com> wrote:
>>>
>>>> Hi Andreas,
>>>> I don't have a good SMARTS pattern available for this but here is a
>>>> function that should return bridgehead idx and not include non bridgehead
>>>> fused ring atoms:
>>>>
>>>> ```
>>>> def return_bridgeheads_idx(mol):
>>>> bh_list=[]
>>>> intersections=[]
>>>> sssr_idx = [set(x) for x in list(Chem.GetSymmSSSR(mol))]
>>>> for i,ring1 in enumerate(sssr_idx):
>>>> for j,ring2 in enumerate(sssr_idx):
>>>> if i>j:
>>>> intersections+=[ring1.intersection(ring2)]
>>>> for iidx in intersections:
>>>> if len(iidx)>2: #condition for bridgehead
>>>> for idx in iidx:
>>>> neighbors = [a.GetIdx() for a in
>>>> mol.GetAtomWithIdx(idx).GetNeighbors()]
>>>> bh_list+=[idx for nidx in neighbors if nidx not in iidx]
>>>> return tuple(set(bh_list))
>>>> ```
>>>>
>>>> Here are 6 test molecules:
>>>>
>>>> ```
>>>> mol1 = Chem.MolFromSmiles("C1CC2CCC1C2")
>>>> mol2 = Chem.MolFromSmiles("C1CC2C1C1CCC2C1")
>>>> mol3 = Chem.MolFromSmiles("N1(CC2)CCC2CC1")
>>>> mol4 = Chem.MolFromSmiles("C1CCC12CCCCC2")
>>>> mol5 = Chem.MolFromSmiles("C1CC2C1CCCCC2")
>>>> mol6 = Chem.MolFromSmiles("C1CCC(C(CCC3)C23)C12")
>>>> for mol in [mol1,mol2,mol3,mol4,mol5,mol6]:
>>>> print(return_bridgeheads_idx(mol))
>>>> ```
>>>>
>>>> giving the expected answer:
>>>>
>>>> (2, 5)
>>>> (4, 7)
>>>> (0, 5)
>>>> ()
>>>> ()
>>>> ()
>>>>
>>>> hope this is helpful!
>>>>
>>>> best wishes
>>>> wim
>>>>
>>>> On Sat, Dec 3, 2022 at 8:34 AM Andreas Luttens <
>>>> andreas.lutt...@gmail.com> wrote:
>>>>
>>>>> Dear users,
>>>>>
>>>>> I am trying to identify bridgehead atoms in multi-looped ring systems.
>>>>> The issue I have is that it can be sometimes difficult to distinguish
>>>>> these
>>>>> atoms from ring-fusion atoms. The pattern I used (see below) looks for
>>>>> atoms that are part of three rings but cannot be bonded to an atom that
>>>>> also fits this description, in order to avoid ring-fusion atoms. The code
>>>>> works, except for cases where bridgehead atoms are bonded to a ring-fusion
>>>>> atom.
>>>>>
>>>>> *PASS:*
>>>>> pattern = Chem.MolFromSmarts("[$([x3]);!$([x3][x3])]")
>>>>> rdkit_mol = Chem.MolFromSmiles("C1CC2CCC1C2")
>>>>> print(rdkit_mol.GetSubstructMatches(pattern))
>>>>> >>>((2,),(5,))
>>>>>
>>>>> *FAIL:*
>>>>> pattern = Chem.MolFromSmarts("[$([x3]);!$([x3][x3])]")
>>>>> rdkit_mol = Chem.MolFromSmiles("C1CC2C1C1CCC2C1")
>>>>> print(rdkit_mol.GetSubstructMatches(pattern))
>>>>> >>>()
>>>>>
>>>>> Any hint on what alternative pattern I could use to isolate true
>>>>> bridgeheads would be greatly appreciated. Maybe other strategies are more
>>>>> suitable to find these atoms?
>>>>>
>>>>> Thanks in advance!
>>>>>
>>>>> Best regards,
>>>>> Andreas
>>>>> _______________________________________________
>>>>> Rdkit-discuss mailing list
>>>>> Rdkit-discuss@lists.sourceforge.net
>>>>> https://lists.sourceforge.net/lists/listinfo/rdkit-discuss
>>>>>
>>>> _______________________________________________
> Rdkit-discuss mailing list
> Rdkit-discuss@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/rdkit-discuss
>
_______________________________________________
Rdkit-discuss mailing list
Rdkit-discuss@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/rdkit-discuss
