Comments on: Standard InChIs and InChIKeys Populated to ChemSpider

By: Tore Eriksson

Tore Eriksson — Fri, 22 May 2009 06:34:23 +0000

Thanks for the replies. Is this what is happening then: In the non-salt (CSID369757) the hydrogens are mobile and the tautamerization makes the sp2 stereoisomers interchangeable. However, if the molecule is protonated the interconversion is stopped/slowed. Or perhaps the stereoisomers only exists in the solid salt (CSID2054355) where interconversion is sterically hindered?

By: Antony Williams

Antony Williams — Sat, 02 May 2009 01:36:17 +0000

CSID2054355 has a crossed bond indicating non-explicit orientation around the double bond…thereby allowing the link on PubChem to both the E and Z orientation. Our experience of OpenBabel is that it is buggy in a number of ways and we prefer to use OpenEye’s tools for this reason.

By: alf

alf — Fri, 01 May 2009 13:23:17 +0000

The difference between PubChem SDF files and PubChem XML-derived SDF files that Tore mentions above seems to be due to OpenBabel, which doesn’t read the PC-Atoms_charge section from the PubChem XML file, so produces an SDF file which doesn’t include the charge information for each atom.

By: alf

alf — Fri, 01 May 2009 13:00:15 +0000

It does seem that the PubChem SDF files that Tore refers to above are missing information that’s present in the corresponding PubChem XML files; the InChIs for those two molecules, using an XML->OpenBabel->SDF->StdInChI conversion, should be:
InChI=1S/C2H8N3S.HI/c1-6-2(3)5-4;/h5H,3-4H2,1H3;1H/b5-2-;
InChI=1S/C2H8N3S.HI/c1-6-2(3)5-4;/h5H,3-4H2,1H3;1H/b5-2+;

I’d be interested to know why this is happening.

By: Tore Eriksson

Tore Eriksson — Thu, 26 Mar 2009 02:29:59 +0000

There semes to be somthing murky with the standard InChI/InChIKeys produced by PubChem. Since I’m not a chemoinformatician, I would appreciate some insight. This might also be relevant for your outlinks. Looking at the PubChem compounds CID:9568014 and CID:15978272 they are stereoisomers but they have identical standard InChI/InChIKeys (NYMQJWVULPQXBK-UHFFFAOYSA-N) without any stereo layer. Searching ChemSpider with the isomeric SMILES or hand-drawn structures both lead to the entry ChemSpider ID:2054355 which also seems to lack sterochemistry. This entry contains outlinks to both CIDs above.

Recomputing the InChI from PubChem SDF files gives the same InChIKey that is shown in all the above entries. However, when I convert the PubChem XML files to SDF (with open babel), and then recompute the standard InChI, the stereo layer is present, and the InChIKeys are of course different. (An extra proton is also added…) I seems like all PubChem InChIs are computed without reagard for stereochemistry now which seems a tad bit odd.

Is this something that you have noticed as well. I am adding InChIKeys to my database now, and wonder if I should merge entries like this.