H-Bond donors/acceptors count using Chemaxon Calculator

User 73ae853b2b

07-12-2009 10:56:38

Hi there,


I encountered
a problem using the ‘Chemaxon Calculator’ component in PP to calculate the
number of H-Bond donors (HBD) and –acceptors (HBA). For a certain reason it was
quite helpful that the component is able to output not only integers but also
the distribution of HBA- and HBD-count at a certain pH. Therefore I am using
the ChemAxon_HBADistribution at pH 7.4 as the number of HBD.


I did some random testing to examine whether this component
gives reasonable results and it seems to work (although I was not able to
figure out every rule how the component classifies one group as an HBD and  another not – is there a documentation about
that?).


But just
today I experienced a very odd situation, where I found a molecule (see
attachment), for which the component calculated 32 HBD, whereas I have counted
only 8 HBD.


I don’t
really see where the 32 HBD come from. Any guesses?

User 73ae853b2b

07-12-2009 12:24:34

I might have found another issue with the Calculator component when I tried to understand how it works. I started with molecule 1 (see attachment). According to the component molecule 1 has three HBA (at pH 7). Subsequently I removed one of the oxygen atoms from the carboxylate group to see how many HBA do account for this particular (removed) atom (molecule 2). Molecule 2 now has only 2 HBA left. To be sure that none of the heteroatoms in the thiazole ring account for an HBA, I replaced the thiazole ring with a phenyl ring (molecule 3). Molecule 3 has the same number of HBA as molecule 2. So, this would mean the carbonyl oxygen formally accounts for two HBA, whereas the negatively oxygen accounts for one HBA.


But then I had a look at molecule 4 – it has only two HBA and I don't really see why?

ChemAxon 9c0afc9aaf

07-12-2009 16:32:30

Hi Simon,


 


My colleagues will answer soon to the calculation specific questions.


Meanwhile could you let us know the JChem version you are using ?


(so we can talk about the same algorithm)


In PP in the Components window under ChemAxon -> Utilities you can find a "Show ChemAxon Environment" component (actually it's a protocol).


The output should contain all relevant version information.


Best regards,


Szilard

User 73ae853b2b

07-12-2009 16:51:39

Hi Szilard,


sorry for not specifying this information. This the output:


ChemAxon For Pipeline Pilot System Information
  JChem version : 5.2.4
Component collection version : 1.5.1
JVM: Sun Microsystems Inc.  Java HotSpot(TM) Server VM  1.5.0_07

------------------------------------------------------------------


 


Cheers,


Simon

User 73ae853b2b

07-12-2009 17:00:43

I noticed something, which might actually turn out to be helpful to trace down the bug: As said before, the 'normal' structure gives an erroneous number of HBD (see 'aromatic.png').


After removing one double bond in the indole ring, the count of HBD is reduced from 32 to 8 (see 'non-aromatic.png')


 


Simon

User 851ac690a0

07-12-2009 19:01:34

Hi,


This bug can occure  in case of  symmetric molecules sometimes. I fix it asap.


Jozsi

User 73ae853b2b

07-12-2009 19:58:51

Hi Jozsi,


thanks for your reply. I thought it must have something to do with the symmetry in this molecule, since adding a methyl group fixes the problem, too!


What about the issue mentioned for molecule 4 in my second post. Am I wrong or should molecule 4 have three HBA?

User 851ac690a0

07-12-2009 23:02:36

Hi,


What about the issue mentioned for molecule 4 in my second post. 
Am I wrong or should molecule 4 have three HBA?



   You are rigth there must be some additional bugs. We fix these bugs asap.


   I attached a figure about the donor/acceptor site counts  of the above mentioned 4 molecules.


   Keto group has 2 acceptor sites and  the "O-" anion also has 2 acceptor sites.


   I considered the "lone pair electron of nitrogen"  as acceptor site, as well.


Jozsi

User 73ae853b2b

08-12-2009 13:54:16

Thanks for your reply Jozsi – makes sense now!


Just mail me if you'd be interested in more symmetric compounds for debugging. I've got a complete list of compounds having a four times higher HBD-count than I would expect, probably because of their symmetry.


Cheers,


Simon

User 851ac690a0

09-12-2009 01:44:13

Hi,


Please attach  a couple of  symmetric structures which were badly handled by the 'HBDA' calculator.    


Thank you in advance!


Jozsi

User 73ae853b2b

09-12-2009 09:35:49

Hi Jozsi,


I attached some badly calculated compounds. I deliberately omitted more symmetric compounds and added some badly calculated non-symmetric compounds instead. (Please note: not all of the stereobonds depicted are correct).


If you're looking for more: here's a list of potentially(!) badly calculated compounds:


ZINC14644092
ZINC13943001
ZINC04014543
ZINC01683973
ZINC03132859
ZINC06021243
ZINC30889031
ZINC03351339
ZINC03351337
ZINC14944886
ZINC17217260
ZINC01632210
ZINC04660209
ZINC03340372
ZINC01708984
ZINC04660212
ZINC30825653
ZINC30825659
ZINC30825650
ZINC01628804
ZINC23245212
ZINC04660213
ZINC21589847
ZINC01656830
ZINC01708990
ZINC30825642
ZINC30825649
ZINC05013559
ZINC04377411
ZINC04377411
ZINC04377411
ZINC04377411
ZINC04039504
ZINC04560214
ZINC04953250
ZINC04953428
ZINC30825654
ZINC30825658
ZINC03869959
ZINC03869958
ZINC03869960
ZINC31520196
ZINC04660121
ZINC05208281
ZINC13298752
ZINC31414378
ZINC06093833
ZINC03873183
ZINC06494177
ZINC13298421
ZINC11326236
ZINC03873187
ZINC03873186
ZINC31769440
ZINC04899463
ZINC04978458
ZINC26329799
ZINC31416748
ZINC31820825
ZINC13578388
ZINC13578388
ZINC13578388
ZINC13578388
ZINC27200745
ZINC27200740
ZINC06092085
ZINC22445959


dig further

User 851ac690a0

10-12-2009 16:54:59

Hi,


 


Thank you for the information.


I fixed the bug that is related with the bad HBDA prediction. 


Fixed version will appear in 5.3


Jozsi