Different results depending on the format

I encounter a really strange problem

I have two chemical compounds, in 2 different databases, that are supposed to be the same:

- kegg:C03516 http://www.genome.jp/dbget-bin/www_bget?C03516

- CHEBI:15431 http://www.ebi.ac.uk/chebi/searchId.do?chebiId=CHEBI:15431

 

If I load these two compounds with MolImporter using their respective molfiles ( http://www.genome.jp/dbget-bin/www_bget?-f+m+compound+C03516 for kegg:C03516
and http://www.ebi.ac.uk/chebi/saveStructure.do;jsessionid=4803C46E7EC8609A093ABB7548715337?defaultImage=true&chebiId=15431&imageId=0 for CHEBI:15431) and ask for the InchI and the SMILES of each, jchem return the same InchI and the same SMILES for both, which is correct :

InChI returned for kegg:C03516

InChI=1S/C34H34N4O4.Mg/c1-7-21-17(3)25-13-26-19(5)23(9-11-33(39)40)31(37-26)16-32-24(10-12-34(41)42)20(6)28(38-32)15-30-22(8-2)18(4)27(36-30)14-29(21)35-25;/h7-8,13-16H,1-2,9-12H2,3-6H3,(H4,35,36,37,38,39,40,41,42);/q;+2/p-2/b25-13-,26-13-,27-14-,28-15-,29-14-,30-15-,31-16-,32-16-;    

SMILES returned for kegg:C03516

CC1=C(CCC(O)=O)C2=N\C\1=C/C1=C(C)C(C=C)=C3\C=C4/N=C(/C=C5\N([Mg]N13)/C(=C\2)C(CCC(O)=O)=C5C)C(C=C)=C4C

InChI returned for chebi:15431   

InChI=1S/C34H34N4O4.Mg/c1-7-21-17(3)25-13-26-19(5)23(9-11-33(39)40)31(37-26)16-32-24(10-12-34(41)42)20(6)28(38-32)15-30-22(8-2)18(4)27(36-30)14-29(21)35-25;/h7-8,13-16H,1-2,9-12H2,3-6H3,(H4,35,36,37,38,39,40,41,42);/q;+2/p-2/b25-13-,26-13-,27-14-,28-15-,29-14-,30-15-,31-16-,32-16-;    

SMILES returned for chebi:15431

CC1=C(CCC(O)=O)C2=N\C\1=C/C1=C(C)C(C=C)=C3\C=C4/N=C(/C=C5\N([Mg]N13)/C(=C\2)C(CCC(O)=O)=C5C)C(C=C)=C4C

 

But now, if I ask for the majormicrospecies at pH 7.3, I have different results depending wether I used the kegg or the chebi molfile, which is strange:

InChI returned for major microspecies of kegg:C03516 at pH 7.3

InChI=1S/C34H34N4O4.Mg/c1-7-21-17(3)25-13-26-19(5)23(9-11-33(39)40)31(37-26)16-32-24(10-12-34(41)42)20(6)28(38-32)15-30-22(8-2)18(4)27(36-30)14-29(21)35-25;/h7-8,13-16H,1-2,9-12H2,3-6H3,(H4,35,36,37,38,39,40,41,42);/q;+2/p-4/b25-13-,26-13-,27-14-,28-15-,29-14-,30-15-,31-16-,32-16-;    

SMILES returned for major microspecies of kegg:C03516 at pH 7.3

CC1=C(CCC([O-])=O)C2=N\C\1=C/C1=C(C)C(C=C)=C3\C=C4/N=C(/C=C5\N([Mg]N13)/C(=C\2)C(CCC([O-])=O)=C5C)C(C=C)=C4C

InChI returned for major microspecies of chebi:15431 at pH 7.3

InChI=1S/C34H34N4O4.Mg/c1-7-21-17(3)25-13-26-19(5)23(9-11-33(39)40)31(37-26)16-32-24(10-12-34(41)42)20(6)28(38-32)15-30-22(8-2)18(4)27(36-30)14-29(21)35-25;/h7-8,13-16,36-37H,1-2,9-12H2,3-6H3,(H,39,40)(H,41,42);/q-2;+2/p-2/b25-13-,26-13-,27-14-,28-15-,29-14-,30-15-,31-16-,32-16-; 

SMILES returned for major microspecies of chebi:15431 at pH 7.3

CC1=C2NC(\C=C3/N4[Mg]N5\C(=C/2)C(C)=C(C=C)\C\5=C\C2=C(C)C(C=C)=C(N2)\C=C4\C(C)=C3CCC([O-])=O)=C1CCC([O-])=O

Any explanation to this?

 

Also, even more strange. If instead of importing the kegg compound from the molfile, I import it using its InChI and ask jchem to give me back the InChI of this molecule, I have different values:

InChI used for the import:

InChI=1S/C34H34N4O4.Mg/c1-7-21-17(3)25-13-26-19(5)23(9-11-33(39)40)31(37-26)16-32-24(10-12-34(41)42)20(6)28(38-32)15-30-22(8-2)18(4)27(36-30)14-29(21)35-25;/h7-8,13-16H,1-2,9-12H2,3-6H3,(H4,35,36,37,38,39,40,41,42);/q;+2/p-2/b25-13-,26-13-,27-14-,28-15-,29-14-,30-15-,31-16-,32-16-;   

InChI returned by jchem:

InChI=1S/C34H34N4O4.Mg/c1-7-21-17(3)25-13-26-19(5)23(9-11-33(39)40)31(37-26)16-32-24(10-12-34(41)42)20(6)28(38-32)15-30-22(8-2)18(4)27(36-30)14-29(21)35-25;/h7-8,13-16H,1-2,9-12H2,3-6H3,(H4,35,36,37,38,39,40,41,42);/q;+2/p-2/b25-13-,26-13?,27-14?,28-15-,29-14-,30-15?,31-16-,32-16?;  

So why is some information, encoded in the InChI used for the import, lost during the process?

SMILES returned by jchem:

 [Mg++].[H]\C1=C2\[N-]\C(=C([H])/C3=N/C(=C([H])\C4=N\C(=C([H])/C5=C(C=C)C(C)=C1N5)\C(C)=C4CCC(O)=O)/C(CCC([O-])=O)=C3C)C(C)=C2C=C

which differ from the one obtained when imported from the molfile...

 

And if now I ask for the major microspecies of this compound (the one imported from the InchI) at pH 7.3 , the information of the Mg is lost...:

InChI returned by jchem for the major microspecies at pH 7.3:

InChI=1S/C34H34N4O4/c1-7-21-17(3)25-13-26-19(5)23(9-11-33(39)40)31(37-26)16-32-24(10-12-34(41)42)20(6)28(38-32)15-30-22(8-2)18(4)27(36-30)14-29(21)35-25/h7-8,13-16,35-36H,1-2,9-12H2,3-6H3,(H,39,40)(H,41,42)/p-2  

SMILES returned by jchem for the major microspecies at pH 7.3:

CC1=C(CCC([O-])=O)C2=NC1=CC1=C(C=C)C(C)=C(N1)C=C1NC(=CC3=NC(=C2)C(CCC([O-])=O)=C3C)C(C)=C1C=C

 

the same kind of inconsistencies occur if I import from the SMILES

SMILES used for the import:

CC1=C(CCC(O)=O)C2=N\C\1=C/C1=C(C)C(C=C)=C3\C=C4/N=C(/C=C5\N([Mg]N13)/C(=C\2)C(CCC(O)=O)=C5C)C(C=C)=C4C

InChI returned by jchem:

InChI=1S/C34H34N4O4.Mg/c1-7-21-17(3)25-13-26-19(5)23(9-11-33(39)40)31(37-26)16-32-24(10-12-34(41)42)20(6)28(38-32)15-30-22(8-2)18(4)27(36-30)14-29(21)35-25;/h7-8,13-16H,1-2,9-12H2,3-6H3,(H4,35,36,37,38,39,40,41,42);/q;+2/p-2/b25-13?,26-13-,27-14-,28-15-,29-14?,30-15?,31-16?,32-16-;    

SMILES returned by jchem:

CC1=C(CCC(O)=O)C2=N\C\1=C/C1=C(C)C(C=C)=C3\C=C4/N=C(/C=C5\N([Mg]N13)/C(=C\2)C(CCC(O)=O)=C5C)C(C=C)=C4C

this time the SMILE used for the import and the one returned are coherent

InChI returned by jchem for the major microspecies at pH 7.3:

InChI=1S/C34H34N4O4.Mg/c1-7-21-17(3)25-13-26-19(5)23(9-11-33(39)40)31(37-26)16-32-24(10-12-34(41)42)20(6)28(38-32)15-30-22(8-2)18(4)27(36-30)14-29(21)35-25;/h7-8,13-16H,1-2,9-12H2,3-6H3,(H4,35,36,37,38,39,40,41,42);/q;+2/p-4  

SMILES returned by jchem for the major microspecies at pH 7.3:

CC1=C(CCC([O-])=O)C2=CC3=C(CCC([O-])=O)C(C)=C4C=C5N=C(C=C6N([Mg]N34)C(=CC1=N2)C(C)=C6C=C)C(C)=C5C=C    
this SMILES differ from the one obtained when importing from the molfile and asking for the major microspecies at pH 7.3 :

CC1=C(CCC([O-])=O)C2=N\C\1=C/C1=C(C)C(C=C)=C3\C=C4/N=C(/C=C5\N([Mg]N13)/C(=C\2)C(CCC([O-])=O)=C5C)C(C=C)=C4C

 

So my question is, why did the results differ depending on the format used for the import?

 

PS: I use jchem release 5.12

Dear Thomas,

I have tried to reproduce your problem, but I could not, because it seems that the two mol files are different. The one from CHEBI database contains coordinate bonds and the other from kegg does not. Major microspecies can not even be calculated for the molecule from the CHEBI database because of the coordinate bonds. Could you please check the two molecules in the mol files?

The SMILES formats does not differ because SMILES does not store coordinate bonds.

Regards,

Zsuzsa

Dear Zsuzsa

       I may have not express myself properly. I agree that the 2 molfiles are not exactly the same as the one from CHEBI contains coordinates bonds that the one from KEGG do not include. However, what I do not understand is why these two mol files give exactly the same InChI but not the same major tautomer at ph 7.3 (the InchIs of the major tautomers differ)? (Remark: the coordinate bonds in CHEBI do not prevent the MajorMicrospeciesPlugin to run and give a result)

The second thing I don't understand is why, Importing the KEGG compound from its InChI gives different results than importing it from its molfile? And especially, why the InChI returned by Jchem is not the same as the one used to import the molecule?

    Thanks for your answers

Regards,

     Thomas

 

 

 

Dear Thomas,

I have checked the issues you have raised. I have found the followings that may cause the problems you have experienced.

• Bonds to metals are stored in the 'AuxInfo' tag of InChI. Without it all bonds to Mg are lost. Is it possible that in some cases  where you used InChI it contained the AuxInfo and in some cases it did not?

•  In InChI coordinate bonds are converted to single bonds. This seems to be a bug, coordinate bonds should rather be omitted than converted to single bonds (as it is done in SMILES), we will fix this issue. This happens with the bonds connected to the Mg as well if AuxInfo is included.

• The facts above mean that the molecule stored in InChI is not the same as in mol beacuse coordinate bonds are lost  both with or without the AuxInfo tag.

•  I have talked to our calculation plugin team about the major miscrospecies problem: the microspecies plugin does not support molecules with coordinate bonds at present, the result can be invalid for a molecule with coordinate bonds. The MajorMicrospeciesPlugin class contains a method checkMolecule(Molecule) which checks the given molecule - whether it is a valid input for microspecies calculation. It raises a PluginException if microspecies calculation is not supported for the given molecule, e.g. if it contains coordinate bonds. By calling this method before calculation you can ensure that microspecies are calculated only for supported molecules. This method is called from MarvinSketch as well that is why I have told you in my earlier reply that microspecies can not be calculated for molecules with coordinate bonds. I admit it is not the best API, the calculation team is planning to solve this problem in the near future as well as supporting coordinate bonds in calculations.

•  I have tried to reproduce your problem with InChI import-export-reimport, but could not reproduce it: the InChI string remained the same.

I hope this helps, best regards,

Zsuzsa

Dear Thomas,

Our next major release, 6.3 will contain the fix. InChi export will not export coordinate bonds as single one, but omit it.

Best Regards,
Peter