Parity

Hello again...





Sorry for bothering everyone again. But I observed some strange behaviour and I am not sure if I am doing it wrong...





I have the following molecule:





C[CH:1](N)[CH:2](C)[CH:3](C)N





I mapped the three stereocenters as #1, #2 and #3. For this molecule I expect 4 stereoisomers (see the 3D-structures in the attached file).


To get those stereoisomers I have to set the parity for the atoms mapped as #1, #2 and #3 as follows (+ means ODD, - means EVEN):





---


+--


++-


+-+





That should give the correct stereoisomers. I checked that the method call mol.setParity(...) returns true in all cases, but after I set the parities only #1 and #3 are carrying the StereoConstants.PARITY_EVEN/ODD flags. That means, despite the fact that the setting of the parity on atom #2 is successful, it is not set at all. Is that due to the fact that #2 is not a asymmetric carbon? And is there a possibility to set the parity nevertheless?





Thanks a lot,





tobias

Hi,





The current Marvin version doesn't support topographically different atoms, just constitutionally different atoms, that's why the middle atom cannot have parity. You set some parity for the middle atom, but as the connected atoms has the same neighbours, the middle atom cannot have a parity.





In Marvin 4.0 this situation has been changed completely. Every atom that has 4 ligands (including implicit H) and has less than two hydrogen atoms can have a parity. This means that you can even assign parity to the carbon in carbon tetrachloride.





All the best


Andras

Hello...





Just another comment on the whole parity thing.


I realized that it is also not possible to set parity on trivalent atoms like N.


Of course I know that N is under normal circumstances not a stereocenter since it is flexible enough to invert but for C1CN2CNC1C2 I would expect two different stereoisomers.


I do not know if this is possible in the new released Marvin 4.0... but I will check that...





tobias

Hi Tobias,





You are right. This is actually not in the marvin 4.0 but sounds pretty good to implement it in the near future.


I think if all ligands of the N is in ring and N is also in the same rings then the N can have chirality (parity). I just wonder if there are more cases when N can have chirality (parity).


Do you know other cases?





Andras

Well, as far as I know a N can only be a stereocenter if it is not free invertible. And this is the case for bicyclic compounds and N at a bridgehead...





tobias

Quote:

bicyclic compounds and N at a bridgehead

I think programmatically this is the same as


"if all ligands of the N is in ring and N is also "





Andras

Yeah, that is true... I am a bit confused now

Quote:

I think if all ligands of the N is in ring and N is also in the same rings then the N can have chirality (parity).

Do you mean with this sentence that it is possible to set the even/odd parity flag for a nitrogen or do you mean, that this should be implemented in a future release. For the first case: I tried it and it was not working, but maybe I did it wrong.


Anyway, I think the whole parity thing is tricky but the two major problems at the moment are:





- no possibility to set parity for carbons with 3 constitutionally ligands


- no possibility to set parity on trivalent atoms (e.g. N)





tobias

volfi wrote:

Quote: bicyclic compounds and N at a bridgehead I think programmatically this is the same as "if all ligands of the N is in ring and N is also "

I guess it is not enough, see this molecule: C1CCC(CC1)N2CCCCC2





Probably you need that all bonds of N are in rings, too.

Hi


This molecule cannot have chirality as it is symmetrical:


C1CCC(CC1)N2CCCCC2


but let us take the following one:


CC1CCCN(C1)C2CCCCC2


I suppose this N cannot have chirality, am I right?


If so, my rule was not right...





Andras

Actually this is not yet in any release, but as soon as the code gets the final version I let it go out. I try to do it asap.





Andras

You are right, the molecule you quoted is the correct counter-example.





Anyhow, "ring N with only ring bonds" would work, wouldn't it?





Szabolcs