Skip to content

dissenting representation of atropisomers #108

Closed

Description

@nbehrnd

Since there is an invite by the InChI project to probe the revised processing of atropisomers, I drew this site's entry 93, M-Binol, on the project's test page editor from scratch:

Image

By click on the molecule editor's floppy disk icon, I'm able to to collect an Extended SMILES string reading

C1(C(O)=C[C@H]=C2C=CC=CC=12)C1C2C=CC=CC=2C=CC=1O |wU:0.0,4.3,(5.43,-3.30,;6.29,-2.80,;7.16,-3.29,;6.29,-1.80,;5.42,-1.30,;4.56,-1.80,;3.69,-1.30,;2.83,-1.80,;2.83,-2.80,;3.70,-3.30,;4.56,-2.80,;5.43,-4.30,;4.56,-4.80,;3.68,-4.31,;2.82,-4.82,;2.83,-5.81,;3.70,-6.31,;4.56,-5.80,;5.43,-6.30,;6.29,-5.80,;6.29,-4.80,;7.16,-4.30,)|

When I paste this string into the the demo page of CDK Depict, the result

  • appears to be unable to add a CIP label - maybe $P, M, \Lambda{}, \Delta{}$ is beyond the intended / requested scope here

  • more importantly: the depiction is puzzling because i) solid wedges are substituted by dashed wedges while ii) retaining the position and the orientation ("wide end" and "apex") of the wedges. On first sight I thought CDK Depict were to depict the enantiomer:

    Image

I'm not sure if this difference is due to CDK depict, or not; in part because the editor as used over on the InChI website (Ketcher version 3.17.0) equally exports the same string as Daylight SMILES.

Activity

  1. johnmay commented on Oct 17, 2026

    @johnmay
    Member

    Looks like a bug with the coordinates getting in the way and wiping out stereochemistry. We actually support the RDKit method so you can do it without coordinates:

    C1(C(O)=CC=C2C=CC=CC=12)C1C2C=CC=CC=2C=CC=1O |wU:0.0|
    C1(C(O)=CC=C2C=CC=CC=12)C1C2C=CC=CC=2C=CC=1O |wD:0.0|
    

    The encoding is described in the RDKit book.

    Also note your input is wrong anyways as you have too many wedges... the sketcher doesn't understand what you have drawn and added an extra tetrahedral stereo which is wrong.

    Image
  2. johnmay commented on Oct 17, 2026

    @johnmay
    Member

    The issue you is you are mixing worldviews. SMILES encodes stereochemistry in it's notation and not the coordinates/display. However atropisomers are not possible in plain SMILES (or even CXSMILES). This will work if you use a MOLfile since stereochemistry there is always captured via the coordinates. CDK Depict supports MOLfile / SDFfiles and so I would recommend that if you have coordinates.

    Coordinates are useful in CXSMILES but it does raise a problem. What if the coordinates disagree with what is captured. For example:

    C/C=C/C |(0.5,.5,;1.2,0,;2,0,;2.5,-.5,)|
    C/C=C\C |(0.5,.5,;1.2,0,;2,0,;2.5,-.5,)|
    

    Gives the little odd:

    Image

    Are the coordinates or the main encoding to be trusted?

    I will look at why the wedge gets flipped around and I think the fix that will reperceive stereochemistry if there is a wedge pointing to an atom which isn't known to be tetrahedral. Reperceiving stereochemistry is quite expensive so I'm not sure I will do it in the toolkit but perhaps the webapp should cross-check things.

  3. nbehrnd commented on Oct 21, 2026

    @nbehrnd
    ContributorAuthor

    The two solid wedges: The lesson is learned; for one, to use only one next to
    the centre or the axis of chirality to be compatible with the software. Though,
    on the other hand, multiple solid wedges on the same ring to visually highlight
    the orientation it is a recurrent motif in textbooks/teaching, discussions,
    publications (e.g.1,2,3,4,5,6) on atropisomerism (and even if there
    is not, illustration of GR-4.3.6 in [Brecher2008]. But this second use is for a
    visual.


    Regarding coordinates: Reaching out for the extended SMILES instead of
    Daylight SMILES was intentional because

    • Chemaxon describes them as part of the string (7)
    • they are mentioned in the RDKit Book(8) (not the cookbook) as accepted input
      and output

    Indeed, a Python script to RDKit (2026.9.1) can export this type of SMILES; run
    on a .mol about methane to report C |(1.0723,-0.0225,-0.0653)|, and on VS057
    of the CIP validation suite a

    O=C(O)c1cccc([N+](=O)[O-])c1-c1c(C(=O)O)cccc1[N+](=O)[O-] |(-1.2471,-1.8247,-0.8726;-2.2873,-1.1891,-0.8817;-3.4299,-1.7023,-0.3967;-2.4457,0.2195,-1.366;-1.3499,1.0577,-1.1013;-1.345,2.3807,-1.5171;-2.4263,2.8835,-2.2226;-3.5201,2.0502,-2.5273;-4.5774,2.6641,-3.3545;-5.7506,2.5119,-3.0007;-4.2188,3.3153,-4.341;-3.5833,0.6783,-2.1074;-4.6905,-0.2467,-2.4925;-4.4681,-1.2882,-3.4443;-3.1118,-1.5324,-4.0132;-2.3459,-0.6886,-4.4464;-2.7682,-2.8264,-3.8819;-5.4908,-2.1555,-3.8593;-6.7703,-2.0324,-3.3447;-7.0341,-1.0748,-2.3767;-6.0134,-0.2078,-1.9258;-6.3733,0.7132,-0.8082;-7.5701,0.9786,-0.6452;-5.4636,1.1337,-0.0829)|
    

    which basically relies on

    from rdkit import Chem
    
    mol = Chem.MolFromMolFile("input.mol")
    cxsmiles = Chem.MolToCXSmiles(mol)

    The 3D .sdf from the site is processed by CDK depict including an assigned M; at
    present, this is not possible with the extended SMILES.

    2026-10-21_data.zip

  4. johnmay commented on Oct 21, 2026

    @johnmay
    Member

    I think you didn't fully understand what I said. There is a slight bug but it does work with CXSMILES just not if you give it coordinates - that is what is confusing it.

  5. johnmay commented on Oct 21, 2026

    @johnmay
    Member

    Also responding here as nothing to do with InChI but you can give CDK depict an SDF file to display multiple MOLfiles.

  6. johnmay commented on Nov 10, 2026

    @johnmay
    Member

    Not deploy yet but we now attempt to synchronise information on inconsistent input.

    Image

    Using the new warning/error handling red icon tells you are doing something odd.

    Image

    If there is either no stereo provided in the coordinates, or none in the base input it presume the other is correct. In this case, removing the [C@] in the base SMILES means there is no warning.

    Image
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions