I would call it a "feature"... I guess running conformer optimization (i.e. ETKDG, UFF, MMFF94) after the embedding would be a good practice...
> I think I do vaguely remember that InChI gives precedence to 3D coordinates if present over anything else for the determination of stereochemistry. I guess that's why there are inconsistencies [sometimes] when the molecule has been generated from a SMILES instead from a MOL block with 2D or 3D coordinates... Christos Christos Kannas Chem[o]informatics Researcher & Software Developer [image: View Christos Kannas's profile on LinkedIn] <http://cy.linkedin.com/in/christoskannas> On Wed, 19 Dec 2018 at 01:12, Markus Sitzmann <markus.sitzm...@gmail.com> wrote: > I think I do vaguely remember that InChI gives precedence to 3D > coordinates if present over anything else for the determination of > stereochemistry. And I think that is what happens here: the Allchem > embedding of the molecule adds 3D coordinates which are not present for the > original molecule create straight from InChI. Probably the minimization of > the structure during the embedding is “turning around” the stereochemistry > (probably you could have a long discussion whether this is a bug or a > feature), > > Markus > > ------------------------------------- > | Markus Sitzmann > | markus.sitzm...@gmail.com > > On 18. Dec 2018, at 19:43, Jason Biggs <jasondbi...@gmail.com> wrote: > > see https://github.com/rdkit/rdkit/issues/1852, and > https://sourceforge.net/p/rdkit/mailman/message/36309813/ > > You can see it in the smiles if you remove stereo after embedding, then > re-detect stereo from the conformation. > > inchi1 = > "InChI=1S/C20H26O4/c1-12(2)17-11-18(22)14(4)7-5-6-13(3)8-16(21)9-15-10-19(17)24-20(15)23/h6-7,10,12,17,19H,5,8-9,11H2,1-4H3/b13-6-,14-7+/t17-,19-/m1/s1" > m1 = Chem.MolFromInchi(inchi1) > m1 = Chem.AddHs(m1) > m2 = Chem.Mol(m1) > AllChem.EmbedMolecule(m2) > m3 = Chem.Mol(m2) > Chem.rdmolops.RemoveStereochemistry(m3) > Chem.rdmolops.AssignStereochemistryFrom3D(m3) > sm1 = Chem.MolToSmiles(m1) > sm2 = Chem.MolToSmiles(m2) > sm3 = Chem.MolToSmiles(m3) > print(sm1 == sm2) # returns true > print(sm2 == sm3) # returns false > > > The difference between sm2 and sm3 is just swapping a \ for a /, > confirming what Christos was able to read from the InChI. > > Why does the inchi reflect the 3D bond stereo but the smiles doesn't until > you remove and re-detect the stereo? Does the InChI code go to the 3D > structure when present and ignore stereo information in the mol object? > > Jason Biggs > > > On Tue, Dec 18, 2018 at 12:14 PM Christos Kannas <chriskan...@gmail.com> > wrote: > >> Hi Jean-Marc, >> >> There difference is due to bond orientation (if my inchi analysis skills >> are correct). >> See the bold bond layer below (14-7+ vs 14-7-). >> >> >> m1 -> >> InChI=1S/C20H26O4/c1-12(2)17-11-18(22)14(4)7-5-6-13(3)8-16(21)9-15-10-19(17)24-20(15)23/h6-7,10,12,17,19H,5,8-9,11H2,1-4H3/*b13-6-,14-7+*/t17-,19-/m1/s1 >> >> m2 -> >> InChI=1S/C20H26O4/c1-12(2)17-11-18(22)14(4)7-5-6-13(3)8-16(21)9-15-10-19(17)24-20(15)23/h6-7,10,12,17,19H,5,8-9,11H2,1-4H3/*b13-6-,14-7-*/t17-,19-/m1/s1 >> >> >> Not sure why it happens, but I've seen it multiple times... >> >> >> Best, >> >> Christos >> >> Christos Kannas >> >> Chem[o]informatics Researcher & Software Developer >> >> [image: View Christos Kannas's profile on LinkedIn] >> <http://cy.linkedin.com/in/christoskannas> >> >> >> On Tue, 18 Dec 2018 at 17:36, JEAN-MARC NUZILLARD < >> jm.nuzill...@univ-reims.fr> wrote: >> >>> Thank you for your answer but alatis might not be adapted to my current >>> problem. >>> >>> Attempting to understand what was changed by the embedding step I wrote: >>> >>> inchi1 = >>> >>> "InChI=1S/C20H26O4/c1-12(2)17-11-18(22)14(4)7-5-6-13(3)8-16(21)9-15-10-19(17)24-20(15)23/h6-7,10,12,17,19H,5,8-9,11H2,1-4H3/b13-6-,14-7+/t17-,19-/m1/s1" >>> m1 = Chem.MolFromInchi(inchi1) >>> m1 = Chem.AddHs(m1) >>> m2 = Chem.Mol(m1) >>> AllChem.EmbedMolecule(m2) >>> sm1 = Chem.MolToSmiles(m1) >>> sm2 = Chem.MolToSmiles(m2) >>> print(sm1) >>> print(sm2) >>> print(sm1 == sm2) >>> inc1 = Chem.MolToInchi(m1) >>> inc2 = Chem.MolToInchi(m2) >>> print(inc1) >>> print(inc2) >>> print(inc1 == inc2) >>> >>> Molecules m1 and m2 have identical SMILES representations >>> but different InChI representations, which I find odd. >>> >>> All the best, >>> >>> Jean-Marc >>> >>> >>> >>> >>> Le 18/12/2018 00:40, Dimitri Maziuk via Rdkit-discuss a écrit : >>> > On 12/17/18 4:50 PM, JEAN-MARC NUZILLARD wrote: >>> >> Is there any more deterministic procedure than the one of trying until >>> >> success is obtained? >>> >> >>> >> How do I determine the InChI string of a conformer obtained after >>> >> multiple embedding? >>> > >>> > This representation keeps 3D config: http://alatis.nmrfam.wisc.edu/ >>> > >>> > Generally speaking the problem with InChI is that the only *required* >>> > layer is the formula. Therefore *an* InChI string cannot be used to >>> > differentiate conformers, you need the InChI string with all the >>> > relevant layers and all the protons. >>> > >>> > https://www.nature.com/articles/sdata201773 >>> > >>> > _______________________________________________ >>> > Rdkit-discuss mailing list >>> > Rdkit-discuss@lists.sourceforge.net >>> > https://lists.sourceforge.net/lists/listinfo/rdkit-discuss >>> >>> >>> _______________________________________________ >>> Rdkit-discuss mailing list >>> Rdkit-discuss@lists.sourceforge.net >>> https://lists.sourceforge.net/lists/listinfo/rdkit-discuss >>> >> _______________________________________________ >> Rdkit-discuss mailing list >> Rdkit-discuss@lists.sourceforge.net >> https://lists.sourceforge.net/lists/listinfo/rdkit-discuss >> > _______________________________________________ > Rdkit-discuss mailing list > Rdkit-discuss@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/rdkit-discuss > > _______________________________________________ > Rdkit-discuss mailing list > Rdkit-discuss@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/rdkit-discuss >
_______________________________________________ Rdkit-discuss mailing list Rdkit-discuss@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/rdkit-discuss
