R/S bug with short bonds

Paste this drawing with two nearly identical structures into Marvin, and turn on View -> Stereo -> R/S Labels -> All.  

<?xml version="1.0"?>

<cml version="ChemAxon file format v5.10.0, generated by v5.10.0">

<MDocument>

  <MChemicalStruct>

    <molecule molID="m1">

      <atomArray atomID="a1 a2 a3 a4 a5 a6 a7 a8 a9 a10 a11 a12 a13 a14 a15 a16 a17 a18 a19 a20 a21 a22 a23 a24 a25 a26 a27 a28" elementType="C C C C C H H C C C O H H H C C C C C H H C C C O H H H" mrvMap="1 2 3 4 0 9 7 6 5 13 0 12 11 14 1 2 3 4 0 9 7 6 5 13 0 12 11 14" x2="3.0139802029921148 1.5663802029921148 2.2482735363254482 3.443500202992115 3.858771073429736 3.9093119500849145 3.01175640345297 4.377580202992116 4.928433536325448 4.877154177166604 6.388495516331329 6.424608933646793 5.956556714295136 5.915230086612446 11.653890952018822 10.206290952018822 10.888184285352155 12.083410952018822 12.569901266731346 12.54922269911162 11.651667152479677 13.017490952018822 13.568344285352154 13.51706492619331 15.028406265358036 15.064519682673499 14.596467463321844 14.555140835639154" y2="0.5371948364429588 -0.19136516355704103 1.0415681697762924 1.8819415031096258 1.0470374495423749 3.3498036956435007 -1.002803557944389 -0.1540318302237078 1.0133815031096254 -1.6107493197622618 -1.90646431182688 0.6486087239027412 2.1599253229968163 -0.23907790833657527 0.512641277786811 -0.21591872221318886 1.0170146111201446 1.857387944453478 0.9512645658206142 3.325250136987353 -1.0273571166005369 -0.17858538887985564 0.9888279444534775 -1.6353028784184094 -1.9310178704830276 0.6240551652465933 2.135371764340668 -0.2636314669927231"/>

      <bondArray>

        <bond atomRefs2="a1 a2" order="1"/>

        <bond atomRefs2="a1 a5" order="1"/>

        <bond atomRefs2="a1 a7" order="1">

          <bondStereo>W</bondStereo>

        </bond>

        <bond atomRefs2="a1 a8" order="1"/>

        <bond atomRefs2="a2 a3" order="2">

          <bondStereo>C</bondStereo>

        </bond>

        <bond atomRefs2="a3 a4" order="1"/>

        <bond atomRefs2="a4 a5" order="1"/>

        <bond atomRefs2="a4 a6" order="1">

          <bondStereo>W</bondStereo>

        </bond>

        <bond atomRefs2="a4 a9" order="1"/>

        <bond atomRefs2="a9 a8" order="1"/>

        <bond atomRefs2="a8 a10" order="1">

          <bondStereo>W</bondStereo>

        </bond>

        <bond atomRefs2="a8 a14" order="1"/>

        <bond atomRefs2="a9 a12" order="1"/>

        <bond atomRefs2="a9 a13" order="1"/>

        <bond atomRefs2="a10 a11" order="2"/>

        <bond atomRefs2="a15 a16" order="1"/>

        <bond atomRefs2="a15 a19" order="1"/>

        <bond atomRefs2="a15 a21" order="1">

          <bondStereo>W</bondStereo>

        </bond>

        <bond atomRefs2="a15 a22" order="1"/>

        <bond atomRefs2="a16 a17" order="2">

          <bondStereo>C</bondStereo>

        </bond>

        <bond atomRefs2="a17 a18" order="1"/>

        <bond atomRefs2="a18 a19" order="1"/>

        <bond atomRefs2="a18 a20" order="1">

          <bondStereo>W</bondStereo>

        </bond>

        <bond atomRefs2="a18 a23" order="1"/>

        <bond atomRefs2="a23 a22" order="1"/>

        <bond atomRefs2="a22 a24" order="1">

          <bondStereo>W</bondStereo>

        </bond>

        <bond atomRefs2="a22 a28" order="1"/>

        <bond atomRefs2="a23 a26" order="1"/>

        <bond atomRefs2="a23 a27" order="1"/>

        <bond atomRefs2="a24 a25" order="2"/>

      </bondArray>

    </molecule>

  </MChemicalStruct>

</MDocument>

</cml>

The only difference between the two structures is that on the left, the C atom in the center is just slightly closer to the C atom above it than it is on the right, but Marvin gives different configurations for the topmost C atom in the two structures.  I don't know why the R/S calculation is dependent on bond lengths, but it should not be.  It should depend only on bond angles.

Hi Bob,

I have moved your question to the structure representation forum, since the chirality calculation is a core chemical calculation, and the visualization of it depends on the values calculated by the structure representation code.

My colleagues will answer for you soon.

Regards,
Istvan

Hi Bob, 

Thanks for the bug report, we started to investigate it. 

Moving the bridge atom up or down should not change chirality. 

We will inform you when the fix is ready. 

Best regards, 

Peter

Dear Bob,

We have investigated the problem, but it seems that we cannot solve it.

The problem is the following:

Given a stereocenter with four ligands in 2 dimension with one wedge. The interpretation is different if one changes the position of the ligand face to the wedged ligand such a way that it is crossing the line connecting the two other ligands. Demo.png presents this situation. If the explicit Hydrogen atom is crossing the segment connecting the I and Cl atoms then the S stereoconfiguration is changing to R. On the other hand, the stereo interpretation does change if you move the wedged ligand (in this case the carbon atom). You can move the wedged ligand anywhere it will be always interpreted as S chirality.

We have tried to handle your example such a way to convert the wedged non ring ligand to the problematic ring ligand (the atom connected to the two bridgehead atoms). In this way the position of this atom would be independent of the result: chirality R always. But it does not seem to work (see demo2.png). The badly represented structure is depicted on figure A, the correct one is on figure B. The problem is that we cannot find a method which correctly convert the wedge in any case. As you can see on the figures (C-F) the chirality changes if we move the wedge to a bond which is a neighbor of the original wedge or lays between the two neighbor bond (on the opposite of the wedge).

We also tried to convert the wedge to the atom which is between the two bridgehead atom but this position is not well defined for example in case of substituted  bicyclo[2,2,2]octane .

What if instead of looking at the line connecting the two ligands, you look at the space defined by the two stereocenter-to-ligand bonds, extended to infinity?  See figure.

I'm on holiday this week, at first sight it sounds to be a good idea, I will examine this when I'm back.

Hi, 

It is really a nice idea, we will have to check the compatibility issues raised by this solution but looks promising.