stereomatching bug

Consider the following three structures. They are identical to one another. JChem 3.1.3 identifies the first two as identical, but not the third. But the middle C is not a stereocenter, so it shouldn't matter whether the configuration of the attached Me group is indicated.

Hi Bob,





This seems to be a tricky one.





The search uses local parity to help distinguish cases like the two molecules below. (These atoms are also not chiral stereocenters, but the two moieties are definitely different -- one cis and one trans.)





For searching (both substructure and perfect), JChem considers those atoms stereocenters which have local parities, so all wedged Carbons in your case count as stereocenters.





However, we will think through these things again and get back to you next week.





Best regards,





Szabolcs

Szabolcs wrote:

The search uses local parity to help distinguish cases like the two molecules below. (These atoms are also not stereocenters, but the two moieties are definitely different -- one cis and one trans.)

Actually, they *are* stereocenters. The definition of a stereocenter is that switching two ligands gives you a stereoisomer. That is the case here. The "four different groups attached" rule is a rule of thumb, but not a definition.

Szabolcs wrote:

However, we will think through these things again and get back to you next week.

Hi guys,





It's past next week. ;-)





-- Bob

bobgr wrote:

It's past next week. ;-)

Yes, sorry, busy weeks...





Basically a query atom with specified stereo configuration cannot match a target atom without stereo. This allows extra filter based on stereocenters even in symmetric cases. For example, see the attached query and target.





To achieve the behaviour you need, I recommend to make a preprocessing step (standardization) on your molecules that remove the unnecessary wedges. Currently the easiest way to achieve this is to convert the molecule to smiles and back. In the future, we will also introduce a Standardizer task for this.


(http://www.chemaxon.com/jchem/index.html?content=doc/user/Standardizer.html)





(The explanation of the smiles trick: smiles export operates on molecules, so it uses "global parity" that considers symmetries also. The searching often deals with substructures, which is only part of the whole picture, so it cannot dump stereochemistry due to symmetry. This is why it uses local parity.)





Best regards,


Szabolcs

Hi Bob,

 

Since JChem 5.3 the default stereo model is the comprehensive model (for some query and markush tables the local model):

https://www.chemaxon.com/jchem/doc/user/query_stereochemistry.html#stereomodels

So the symmetric targets are matched by the stereo queries.

Regards,

Robert

From what I can see, the comprehensive model does not solve the original problem (see above), either for substructure or exact searches.

INFO: comparetwo.jsp: searchType = exact

comparetwo.jsp: setStereoSearch = true, stereoSrchStr = true

comparetwo.jsp: exactStereoMatching = false

comparetwo.jsp: doubleBondStereoMatching = false

comparetwo.jsp: stereoMatchingModel = comprehensive

comparetwo.jsp: chargeType = exact

comparetwo.jsp: radicalType = exact

comparetwo.jsp: isotopeType = exact

comparetwo.jsp: valenceType = true

comparetwo.jsp: orderSensitive = false

comparetwo.jsp: listAllMatches = false

comparetwo.jsp: target = C[C@@H]1CC[C@@H](C)C1C

comparetwo.jsp: query = C[C@@H]1CC[C@@H](C)[C@H]1C

comparetwo.jsp: searchResult = false

INFO: comparetwo.jsp: searchType = substructure

comparetwo.jsp: setStereoSearch = true, stereoSrchStr = true

comparetwo.jsp: exactStereoMatching = false

comparetwo.jsp: doubleBondStereoMatching = true

comparetwo.jsp: stereoMatchingModel = comprehensive

comparetwo.jsp: chargeType = exact

comparetwo.jsp: radicalType = exact

comparetwo.jsp: isotopeType = exact

comparetwo.jsp: valenceType = true

comparetwo.jsp: orderSensitive = false

comparetwo.jsp: listAllMatches = false

comparetwo.jsp: target = C[C@@H]1CC[C@@H](C)C1C

comparetwo.jsp: query = C[C@@H]1CC[C@@H](C)[C@H]1C

comparetwo.jsp: searchResult = false

Am I missing something?

Hi Bob,

We are checking the molecule (C[C@@H]1CC[C@@H](C)[C@H]1C), why it doesn't match the one without chiral information on the middle atom.

Bye,

Robert