Different InChiKeys for the same molecule

Hello,

Here is an example of a compound for which the Marvin generated InChIKey is different from the one from IUPAC.

c1ccc(cc1)C(N)P(=O)(O)O
IUPAC : ZKFNOUUKULVDOB-XMBMESGPNA-N
MARVIN : ZKFNOUUKULVDOB-UHFFFAOYNA-N

There have been a few other examples that I have not recorded, unformtunately. The problem is, I use inchikeys to retrieve compound-specific information from my databas, and not everybody is using Marvin to generate InChi keys. So, one might not find a molecule that is actually in the database. Does this difference come from possible differences in the normalization of the molecule (assuming it is done prior to generating the inchi)? What could be the bst way to deal with this?

Regards,

Yan

Hi Yan,

As far as I know, the InChi software accepts only MOL or CML files.

What are you using for smiles to MOL conversion? Could you provide the InChi string as well as the InChiKeys?

Thank you for the report.

BRs

Peter

Hi Peter,

I actually have the smiles as input and I use JChem to convert to inchikeys. I was provided the IUPAC inchikey by a scientist using testing my application. In this application, every conversion is done with JChem. I assume I do not have convert first from smiles to MOL. Am I right?

Best Regards,

Yannick

In JChem you may convert to InChi from any of the supported formats (including SMILES), but the IUPAC official software accepts only MOL as input format.

It would be interesting to find out what was the representation of the molecule was when the provided IUPAC key was generated.

I've tried and I could not reproduce the InChiKey you refer as "IUPAC InChi Key" with the IUPAC software. I'm afraid without the origin of the other key we won't be able to get to the bottom of this issue.

Hello,

Actually, I realized that the other tool (inchi-1 from the InChi Trust @ http://www.inchi-trust.org/downloads/)was generating InChiKeys based on non-standard InChIs.

I also tried OpenBabel to see whether I would get the same InChi/InChikey. For the molecule attached here, Both molconvert and babel  returned the same InChi:

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

molconvert inchi:AuxNone ~/test.mol

Omitted undefined stereo

InChI=1/C7H10NO3P/c8-7(12(9,10)11)6-4-2-1-3-5-6/h1-5,7H,8H2,(H2,9,10,11)

babel -isdf ~/test.mol -o inchi

==============================

*** Open Babel Warning  in InChI code

  #1 :Omitted undefined stereo

InChI=1S/C7H10NO3P/c8-7(12(9,10)11)6-4-2-1-3-5-6/h1-5,7H,8H2,(H2,9,10,11)

1 molecule converted

1 warnings 16 audit log messages 

ZKFNOUUKULVDOB-UHFFFAOY(N)A-N

ZKFNOUUKULVDOB-UHFFFAOY(S)A-N

This is a problem when having creating a database resource that could be searched through InChikeys. Pleople could submit InChikeys that do not return any hit alothough the actual molecule is in the database.

I assume that this could happen because of differences in the algorithms for the inchikey generation or th implementations thereof.

Regards,

MrYan

Hi there,

Thank you for the information.

The reason behind the difference in InChiKeys is that the CHIRAL flag is not set in the attached molfile and openbabel may ignore this.

If I put the chiral flag into the molfile, the generated inchikeys are the same. ( It's the "1" in the 4th row in the 4th column).

You can force the usage of absolute stereo in all molecules using the "inchikey:SAbs" export option.

Best regards,

Peter

Thank you Peter.

Regards,

MrYan