stereochemistry problem in Jchem search

User b91dd8facf

15-11-2011 10:14:44

I have created a JChem (version 5.5.1.0, Sun Java HotSpot(TM) 64-Bit Server VM, 1.6.0_13, amd64 SunOS 5.11) database with 1 compound:


O=C(C)O[C@H]1C[C@@]2(C)[C@@H]3CC=C4[C@@H](C[C@H](OC(C)=O)C(=O)C4(C)C)[C@]3(C)C(=O)C[C@]2(C)[C@H]1[C@](O)(C)C(=O)\C=C\C(C)(C)O |c:10|


than, using the supplied jsp dbsearch utility I am using the substructure search string


C[C@@]12CCC[C@@]1(C)[C@@H]1CC=C3[C@@H](C[C@H](O)CC3(C)C)[C@]1(C)CC2


I am getting back the a compound with the following smiles:


O=C(C)O[C@@H]1C[C@]2(C)[C@H]3CC=C4[C@H](C[C@@H](OC(C)=O)C(=O)C4(C)C)[C@@]3(C)C(=O)C[C@@]2(C)[C@@H]1[C@](C)(O)C(=O)\C=C\C(C)(C)O


which is the mirror image of the compound that is in the database and does not correspond to the search string...


Very strange! If I am using the compound cd_id for searching I am getting back the correct smiles...


What is going on here?


best regards


Lutz

ChemAxon a3d59b832c

16-11-2011 10:45:49

Hi Lutz,


 


I tried to reproduce the problem on our demo site (JChem 5.7. Table editexample):


http://www.chemaxon.com/jchem/examples/db_search


 


However, I could not. The coordinates / orientation of the target and query structures may be important, as it seems that something could be wrong during hit alignment.


Could you perhaps attach the structures in e.g. molfiles or mrv files?


 


Thanks,


Szabolcs

User b91dd8facf

16-11-2011 17:40:41

Hi Szabolcs,


indeed, when I try this with the smiles there is no problem. It occurs only with the SDF which I included. The substructure searching was done using the search smiles and copying it into the editor window...


cheers


Lutz

ChemAxon 8407015329

17-11-2011 10:04:27

Hi Lutz,


I verified the search with the query using SMILES and SDF file as input. The cause of the difference is in the input of the query:


- when importing the SMILES the structures are aligned the same way, therefor no flip is performed on the target molecule


- when importing the SDF file i noticed that the query molecule is flipped 180 degrees among the X coordinate which results in flipping the wedges as well. The SDF flie contains the flipped version of the molecule.


The scenario you are experiencing is due to hit alignment, we provide the target molecule transformed to be as close to the query molecules orientation as much as possible. Alignment can switched off in which case no flipping will be performed.


 


Regards,


Vencel

User b91dd8facf

17-11-2011 16:43:10

Hi Vencel,


maybe I could not make myself understand:


I have loaded into a database the SAME molecules twice:


first as a smiles:


O=C(C)O[C@H]1C[C@@]2(C)[C@@H]3CC=C4[C@@H](C[C@H](OC(C)=O)C(=O)C4(C)C)[C@]3(C)C(=O)C[C@]2(C)[C@H]1[C@](O)(C)C(=O)\C=C\C(C)(C)O
|c:10|


second as a MOL file which I did send you, which has the same smiles. The database content is exactly the same for both molecules with the exception of the BLOB and the cd_id.


Than I take the query smiles string


C[C@@]12CCC[C@@]1(C)[C@@H]1CC=C3[C@@H](C[C@H](O)CC3(C)C)[C@]1(C)CC2


and perform a sub-structure search. This returns both molecules but the one which was loaded from the MOL file is shown as its ENANTIOMER !!! moreover, when I copy the smiles behind the display picture, it is also the enantiomer !!! So its not just about flipping the molecule - it is changing its stereochemistry, which is entirely wrong...  This means your alignment algorithm is not only flipping the molecule in space, it is also changing the stereochemistry in certain, unpredictable cases. I would recommend to disable this algorithm completely - as we are currently trying to work on natural products which have a lot of stereo centers this was leading us wrong...


cheers


Lutz

ChemAxon 8407015329

18-11-2011 09:14:22

Hi Lutz,


I followed the steps mentioned and was able to reproduce the issue you experienced, in JChem version 5.5.1. I checked the results in the viewer with stereo/RS labels displayed. The issue has been fixed since then and is available in our newest release 5.7. You can check the new behaviour on our website: http://www.chemaxon.com/ajax/ i added the two structures to the editexample test database table. Is this what you would expect?


To solve the problem you can do two things:


- upgrade to 5.7 is this is conveniant


- disable hit alignment


Regards,


Vencel