Bridged ring system represented incorrectly

Hi ChemAxon folks,

We have the following two structures:

1. CC(C)(C)OC(=O)N1C[C@H]2C[C@@H]1CN2

2. CC(C)(C)OC(=O)N1C[C@@H]2C[C@H]1CN2

When they are displayed by the ChemAxon tools, #1 is represented correctly, but #2 is shown in a misleading and physically impossible configuration. Please see the attached images.

Krishna

Hi Krishna,

This is an issue of displaying stuctures. 

In the  #2 structure, the ring part with the  NH is in the front, and the other one is at the back. It is displayed incorrect because the viewer can't decide which bond is in the front.

Is this causing some trouble for you?

Best regards, 

Peter

Hi Peter,

Yes, it is definitely a problem for our customers that the structure is drawn incorrectly. Here is an example of structure #2 drawn correctly: http://www.chemspider.com/Chemical-Structure.9696050.html

When is this going to be fixed?

Best regards,

Krishna

[EDIT: The structure shown in the chemspider link is also wrong! Please disregard that link.]

The example you showed has two explicit Hydrogens on the bridge atoms. In that case we put the wedge on the Hydrogens too. Here is the two Smiles with expicit Hydrogens: 

1. [H][C@@]12CN(C(=O)OC(C)(C)C)[C@@]([H])(CN1)C2

2. CC(C)(C)OC(=O)N1C[C@@]([H])2C[C@]([H])1CN2

I will ask my collegaues how and when we can improve the display of this kind of structures.

Regards,

Peter

Just to clarify, the explicit hydrogens are not the problem. I included that link as an example of one possible correct representation of the core structure.

Krishna

Hi Krishna,

having problems with the discussion here, I would like to get the 3D-structure of the molecule

http://www.chemspider.com/Chemical-Structure.9696050.html

The 3D button on this page doesn't show the 3D structure. The option to save the molecule produces only a 2D-mol-file:

625.mol
  -OEChem-04041210272D

 14 15  0     1  0  0  0  0  0999 V2000
    4.7258   -3.7713    0.0000 N   0  0  0  0  0  0  0  0  0  0  0  0
    5.2809   -3.6239    0.0000 C   0  0  2  0  0  0  0  0  0  0  0  0
    5.8401   -3.7713    0.0000 C   0  0  0  0  0  0  0  0  0  0  0  0
    6.1288   -3.2743    0.0000 N   0  0  0  0  0  0  0  0  0  0  0  0
    5.0145   -3.2743    0.0000 C   0  0  0  0  0  0  0  0  0  0  0  0
    5.5738   -3.1268    0.0000 C   0  0  2  0  0  0  0  0  0  0  0  0
    5.2824   -2.6255    0.0000 C   0  0  0  0  0  0  0  0  0  0  0  0
    6.6266   -2.9867    0.0000 C   0  0  0  0  0  0  0  0  0  0  0  0
    7.1286   -3.2743    0.0000 O   0  0  0  0  0  0  0  0  0  0  0  0
    6.6266   -2.4117    0.0000 O   0  0  0  0  0  0  0  0  0  0  0  0
    7.6264   -2.9867    0.0000 C   0  0  0  0  0  0  0  0  0  0  0  0
    8.1283   -3.2743    0.0000 C   0  0  0  0  0  0  0  0  0  0  0  0
    7.6264   -2.4117    0.0000 C   0  0  0  0  0  0  0  0  0  0  0  0
    8.1283   -2.6992    0.0000 C   0  0  0  0  0  0  0  0  0  0  0  0
  1  2  1  0  0  0  0
  2  3  1  0  0  0  0
  3  4  1  0  0  0  0
  1  5  1  0  0  0  0
  5  6  1  0  0  0  0
  4  6  1  0  0  0  0
  6  7  1  0  0  0  0
  2  7  1  0  0  0  0
  4  8  1  0  0  0  0
  8  9  1  0  0  0  0
  8 10  2  0  0  0  0
  9 11  1  0  0  0  0
 11 12  1  0  0  0  0
 11 13  1  0  0  0  0
 11 14  1  0  0  0  0
M  END
$$$$

Loading this file in MarvinSketch-5.9.1 shows a "Chiral Flag Error" and "Clean-3D-with Hydrogenize" produces the RR-enantiomer. I would be good to have here your mol-File in 3D.

Best regards, Hans-Ulrich

Hi Krishna,

I understand that the Hydrogens are not the problem, but they are part of the solution.

If you use the option Structure->Clean 2D-> Hydrogenize chiral centers, then the H-s are put on the sturcture and got the proper wedges. You can use this option during batch conversion too:

molconvert mrv -2:H -s "CC(C)(C)OC(=O)N1C[C@H]2C[C@@H]1CN2" 

At the moment you can't convert to smiles format this way, because this is a cleaning option and cleaning is not available for smiles export.

Unfortunately the bonds are drawn out as they are ordered in the molecule. We can't change the drawing order when a bond type is changed. 

Regards,

Peter

Hi Peter,

I just realized that the structure shown in the chemspider link is also wrong! Please disregard that link. Here is an example of the 1S,4S strucure drawn correctly: http://www.sigmaaldrich.com/catalog/product/aldrich/473049?lang=en&region=US (Hopefully that clears up your confusion, Hans-Ulrich).

When I paste the SMILES with the explicit hydrogens into Marvin, the bridged ring system is still drawn incorrectly. Yes the the R/S labels are correct, but the ring system is wrong. I'm not sure how I can state this clearly enough: changing wedges or adding hydrogens does not alter the fact that the bridge is drawn on the wrong side of the ring.

In order to render a correct structure, either the bridge must be drawn below the ring, or the positions of the nitrogens in the ring must be changed.

Krishna

Hi Krishna, let me discuss here another possibilty:

Those bicycloheptane derivatives are normally drawn in the usual "pseudo 3D formula". But it could also be drawn as a 6-membered ring and the methylene bridge inside the ring. Drawing now one bridge down (MDL style), as shown in the first image, it produces the correct SS-structure and MarvinSketch has "No problems". And it seems to me, this formula is the best way to follow the IUPAC recommendations 2005. It's preferred to have wedge bonds not inside a bond angle.

And, as can be seen in the Ball and Stick 3D view, this formula is a good presentation of this structure, too.

This formula gives the SMILES: CC(C)(C)OC(=O)N1C[C@@H]2C[C@H]1CN2, but pasting it back to MarvinSketch produces bold wedge bonds in the methylene bridge having a problem with a bond length.

Best regards, Hans-Ulrich

Dear Krishna,

Is the source of the problem that the depiction breaks two crossing bond?

So in 1.png the bond pointing upward is not broken into two part and so the bold part of the ring is in front.

Am I understand the problem correctly?

Hi Volfi,

Yes, if the bond pointing upward were broken, so that the part of the ring you have drawn in bold was understood to bee in back, that would be (one) correct way to draw it. Probably not the best way to draw it, however. I've attached an image of a nice representation of the S,S molecule.

Hi  Hans-Ulrich,

The drawing method and SMILES you mention are good, but the problem is that ChemAxon will draw a misleading structure when given your SMILES. You can see this if you paste the SMILES into the marvin applet at http://www.chemaxon.com/marvin/sketch/index.php . In my specific case I am generating static images from CXSMILES, but the result is the same as what you see after pasting into the applet.

Best,
Krishna

Hi Krishna,

before continuing the discussion here one question: How do you draw the enantiomer ?

Best regards, Hans-Ulrich

PS: And one more: Is there a trivial name for this compound ? I want to put a 3D presentation on my 3D pages.

Hi Hans-Ulrich,

You mean the 1R,4R enantiomer? That is pictured in the original post in image_1.png.

I'm not aware of a trivial name, unfortunately.

Best regards,

Krishna

Hi Krishna,

yes, I mean the 1R,4R enantiomer. But my question was: How do you draw the 1R,4R enantiomer in a presentation without wedge bonds, just as in the image SS.png ?

Best regards, Hans-Ulrich

Hi Krishna,

Regarding to the broken bonds:

"bond pointing upward were broken, so that the part of the ring you have drawn in bold was understood to be in back"

We don't support stereochemistry without wedge information and the only the wedge information is taken into account to determine the stereo descriptor (the broken bond info which could refer to which bond is the front and which is at the back is not used).

Just take image_1.png and convert the wedge bonds to regular bonds. Right? Is there something I'm missing?

Krishna

I'm not sure what you are trying to say. Let me start back at the beginning:

1. When given the valid SMILES CC(C)(C)OC(=O)N1C[C@@H]2C[C@H]1CN2, the ChemAxon tools draw a structure that is wrong.

2. I would like to know when this will be fixed.

Hi Krishna,

OK. I tried to draw the enantiomer with the Graphics Image Manipulation Program GIMP, see attached image. And I have also tried to draw the mirror reflecting the enantiomers. All these representations are projections of the 3D objects to the drawing plane.

In the next post, I will try to produce the enantiomers with wedge bonds.

And now the trial to produce the enantiomers with "correct" wedge bonds, see attached image.

It seems to me, it's very problematic to mix 2D and 3D in the same formula.

Best regards, Hans-Ulrich

Hi Hans-Ulrich,

Yeah, adding the wedge bonds to the pseudo-3D rendering is awkward, but I believe our customers would be happy with that representation or the one in your previous post. Either representation makes it clear which enantiomer is which.

Krishna

I'm not sure what you are trying to say. Let me start back at the beginning:


1. When given the valid SMILES CC(C)(C)OC(=O)N1C[C@@H]2C[C@H]1CN2, the ChemAxon tools draw a structure that is wrong.


2. I would like to know when this will be fixed.

Currently we use wedge representation of chiral centers and we do not support stereo recognition without wedges even in such cases when it could be detected for a rigid structure with perspective drawing.

To able to cope this case we need to recognize stereocenters from perspective drawing which is not yet supported.

So if you need a representation without wedge bonds like the ones shown at BridgedRingSystems-3.png, it is a huge work. We will definitely have no time for this "fix" (which includes plenty of features) in this year.

The wedge representation solution BridgedRingSystems-5.png could be fixed much easier. Or we can make a clean option to use a representation shown at SS-Bicyloheptane-1.png.

please let me know your opinion

Hi, my opinion is already documented here, but I will repeat it:

1) Using both stereographic projection and wedge bonds in the same molecule drawing is very problematic. These two different representations are two different "languages".

2) The representation  SS-Bicyloheptane-1.png is compatible to the rules about wedge bonds and especially to the IUPAC recommendations 2005.

So I will add both enantiomers using this representation.

And the enantiomers are produced Marvin's Edit - Transformation - Mirror - to the 3 possible planes, see attached image. MarvinSketch works perfectly.  

Best regards, Hans-Ulrich

Hi Volfi,

Thank you, that is very helpful. I do not think we need a perspective-only representation without wedge bonds. Of the other two choices, it sounds like using wedge bonds only (as in SS-Bicyloheptane-1.png) is the preferred option. I've asked our clients if they have any input.

However, I have one question: you mention adding a clean option to achieve the wedge bond-only representation. Does this mean that we would need to know which compounds required cleaning? This would be problematic, since our application must generate images from CXSMILES without any manual intervention. If the clean setting could be safely applied to all structures then that would be fine.

Hans-Ulrich: thank you again for your help with this!

Best regards,

Krishna

I do not think we need a 

perspective-only representation without wedge bonds. Of the other two 

choices, it sounds like using wedge bonds only (as in  SS-Bicyloheptane-1.png) is the preferred option. I've asked our clients if they have any input.

OK we will make the perspective representation optional and solve this problem this way.

However, I have one question: you mention adding a clean option to 

achieve the wedge bond-only representation. Does this mean that we would

 need to know which compounds required cleaning? 

I was not clear enough. "clean" means for me 2D coordinate generation. When you depict a SMILES you need 2D coordinates. That is what I usually call "2D cleaning". You don't need to do anything just import the molecule, create 2D coordinates (2D cleaning) and show the result.

Sorry for the confusion.

Hi Volfi,

Great-- our customer has confirmed that they like the 'wedge bonds only' representation. What is the rough timeline for this fix?

Thanks,

Krishna

We are at the commit deadline of Marvin 5.10 release at Friday, and we cannot fix this bug until then as we have other planned fixes which need to be fixed.

I suppose we can fix this in an 5.10 bugfix release.

Hi Volfi,

Roughly when should we expect a 5.10 bugfix release? Is this forum thread the best place for watch for updates on the status of this bug?

Thanks,

Krishna

I hope in a month or two.

Our policy is that if we fix the bug we will inform the customer. In this case we will write in this forum topic something like "The bug is fixed, the next Marvin release 5.10.X will contain the fix"

Great, thank you Volfi.

Krishna

Hi Volfi,

Since 5.10 has been out for over a month now, I wanted to check on the status of this. Is a bugfix release planned, and if so, roughly when?

Thanks,
Krishna

Hi Krishna,

The fix is planned to 5.10.2.

5.10.1 is on the way and the next release will contain this fix also.

Volfi

Hi Krishna,

In 5.10.2 (coming out soon) new clean option, Toff, is introduced to ignore template based cleaning. 

Using this option the generated image is attached.

Hi Volfi,

I'm glad to hear that a potential solution will be available soon. However, I'm very concerned that this will compromise the display of other structures, where template-based cleaning is critical for proper display. Is this the case? As you would expect, we need a single configuration that provides reasonable display of the vast majority of structures, and for the remaining structures, where the drawing may be poor, we want it to at least be intelligible.

Thanks,

Krishna

Hi Krishna,

You are right, you should choose between template based or non template based cleaning. However I undrestand your request, you want to ignore the templates in some cases.

Could you define in which cases would you use the non template based version.

I feel that you don't want template based cleaning if ring template atom has well defined parity/chirality value. Is that the case?

Hi,

What is the result for the formula above using the Marvin tools "Transfromation - Mirror - Horizontally - Vertically - To Canvas" ?

Best regards, Hans-Ulrich

Hi Volfi,

We need to display a wide variety of structures, representing the full chemical diversity you would expect in drug discovery efforts. We want the non-template-based version in the cases where the template is wrong! As I understand it, the template you use to clean CC(C)(C)OC(=O)N1C[C@@H]2C[C@H]1CN2 has a bug, resulting in a drawing that is wrong (image_2.png in my original post). Maybe you need to fix the template? I wish I could tell you how to solve the problem, but since I am not an expert on the internal workings of your software, I cannot do that.

Krishna

Hi Krishna,

You may misunderstood my question. I don't expect you to tell me how to fix the problem, I just would like to clearly  understand your request. Let's go to the details. Template based cleaning does not use wedges at all (not even specific atom types). It just uses 2D coordinates, so the template we use is the 2D coordinates of the following molecule: C1CC2CCC1C2.The wedges are arranged after all 2D coordinates are generated. So how do I know which template is wrong? How do I know in which cases should ignore the template coordinates? Is the statement I worte

"I feel that you don't want template based cleaning if ring template atom has well defined parity/chirality value."

is correct?

Dear Hans-Ulrich

Should I attach of the resulted molecules as images? Or you wanted to tell that there is a bug somewhere?

Dear Volfi,

I only wanted to see it as image.

Best regards, Hans-Ulrich

Dear Hans-Ulrich,

Here are the images.

Dear Volfi,

looks excellent. Thank you.

It's really good to have no crossing bonds.

And the images with upside solid wedge bonds are very similar to the often used view from side with crossed bonds (for example the MarvinSketch templates).

I will continue to test these systems.

Best regards, Hans-Ulrich

If I can add my two cents' worth...

Although Hans is right that it is problematic to mix 2D and 3D representations, nevertheless, it happens often.  When I teach my students how to draw the products of Diels-Alder reactions, they have to be very careful how to draw their tructure, or Marvin misunderstands it.  The conventional way of drawing the Diels-Alder adduct of cyclopentadiene and acrolein is the first structure, but, as you say, Marvin does not recognize the stereochemistry without wedge bonds.  Students' natural response is to make the two bonds to the bridging CH₂ group bold and the bond to the CHO group hashed, but Marvin now misinterprets the configurations.  The way to draw the structure so that Marvin interprets it correctly is the third structure; but some students will draw the fourth structure instead, and Marvin will misinterpret the configurations (because of the position of the CH₂ group inside the ring).  Ideally, Marvin should be able to recognize all four structures as the same compound, with the configurations denoted in the third structure.  How you write an algorithm to do this, I have no idea.  

[Wednesday evening, I am unable to upload the PNG or JPG picture.  I'll try again tomorrow.]

Hi Bob,

before continuing the discussion here, I have two questions, both concerning the fourth structure:

1) Where do you see the bridging CH2 group, below or above the drawing plane of the six membered ring ?

2) Why are the two bridge head C-H bonds hashed and pointing in the same direction as the formylgroup ?

Best regards, Hans-Ulrich

Let me answer (2) first.  There is a mnemonic device for drawing the stereochemistry of Diels-Alder adducts called the "out-endo -> cis" rule.  (See Figure 1.)  If you consider the shape of the s-cis diene as an arc, the four substituents on the termini of the diene are either on the convex, or outer, side of the arc, or the concave, or inner side of the arc.  We call the groups on the convex face the "out" groups, and those on the concave face the "in" groups.  The  "out-endo -> cis" rule tells you that the out groups on the diene and the endo groups on the dienophile are always cis in the adduct.  The student simply needs to identify which group on the dienophile is going to prefer to be endo in the TS for the Diels-Alder reaction, and all the stereochemistry follows.  You can also apply the "out-endo -> cis" rule to in and exo groups: toggle either of the first two terms in the rule, and the third term toggles also.  So we also have "in-endo -> trans", "out-exo -> trans", and "in-exo -> cis" rules.  

OK, now let's consider the "out-endo -> cis" rule in the context of cyclopentadiene (CpH).  In CpH, the H atoms are the out groups, and the CH₂ group is the in group for both termini of the diene.  If we apply the "out-endo -> cis" rule, the H atoms should be cis to the formyl group in the adduct.  If we apply the "in-endo -> trans" rule, the CH₂ group should be trans to the formyl group in the adduct.  Both rules must lead to the same product!  Yet Marvin treats these two structures as diastereomers.  Interestingly, if the student applies both rules and draws two wedged bonds coming from the bridgehead C atoms, Marvin ignores the wedges to the bridging CH₂ group.  

Currently, I have to give the students workarounds in the case that they are applying the "out-endo -> cis" rule to a cyclic diene.  Either they need to draw the wedges to the in groups only, or they need to assign R and S to their adducts, and they then need to adjust their wedges in their Marvin drawing so that Marvin is assigning the same configurations.   

So, here is what I would like.  In cases where there is a substituent inside the ring, as in the CH₂ group in the examples, I would like the stereochemistry interpreted as if the substituent were actually outside the ring.  I understand that this solution may cause problems in other cases, so it may not be possible.  But it's what I would like. 

Hi Bob,

thank you for your detailed answer. Very interesting and I have already downloaded the OutEndoCisRule graphics.

Now I see the problem: Marvin notifies errors for wedge bonds on different sides of planar bonds, as shown in the attached image. The IUPAC Recommendations 2006 "Graphical Representation of Stereochemical Configuration", Pure Appl. Chem. 78, 10, 1897-1970 characterize this as "Not acceptable". We have already discussed about this paper some time ago.

But I agree to your argumentation: There are situations where in spite of disregarding these rules, the geometry of the molecule is clearly defined, especially for bicyclic systems.

I will further analyse the situation. Perhaps there is a solution. There is just coming an idea ...

Best regards, Hans-Ulrich

Hi Volfi,

First, for bridged rings like this, the template ChemAxon uses always produces a 3-D drawing with implied stereochemistry, even if the SMILES does not have stereochemistry. For example, CC(C)(C)OC(=O)N1CC2CC1CN2 produces the structure seen in "should_not_have_stereochemistry.png". 

Second, as we have seen, the template cleaning does not correctly consider stereochemistry, so the structure that started this whole discussion is drawn incorrectly. As Hans-Ulrich has pointed out, it is problematic to mix 2-D and 3-D representations, so using the template seems like a bad idea when the stereochemistry is defined.

Given these two facts, it seems to me like this particular template should NEVER be used. If you could fix the template itself so that it would draw the mirror image when needed (see bridgedringsystems-3.png), I suppose that would be OK, but based on what Hans-Ulrich has been saying it sounds like 2D-only representations are the way we are going, in which case I do not see the point of this particular template.

Thoughts? Does this make sense?

Krishna

Even if you go to 2D-only templates, there will remain the problem of interpreting structures drawn by students that I raised.  In other words, Krishna and I are concerned about somewhat different, though related, problems.

Dear All,

I understand Bob's problem, but actually I cannot see the solution for it. Ideas are welcome.

I also understand Krishna's problem. The original problem I see that the templates implies some stereo information usually, but not in ChemAxon.  As I wrote before we can detect stereochemistry if wedge is defined at the stereocenter. Now let me understand Krishna's comment: if stereochemistry is not defined in the SMILES (eg:CC(C)(C)OC(=O)N1CC2CC1CN2)  we should not use templates as it is misleading, if stereochemistry is defined in the SMILES we should not use templates either as it is misleading. So we should not use this template at all, but earlier you explicitly mentioned that you don't want to avoid templates "I'm very concerned that this will compromise the
display of other structures, where template-based cleaning is critical
for proper display." So only specific templates should be avoided. Is this the case?

Dear All,

before continuing the discussion here I must clear our nomenclature and I will use Bobs examples, see attached image.

Does my "nomenclature" correspond to the posts here ? If not I will change it.

And now to the stereochemistry: Often I use Clean 3D Fine with Hydrogenize to check the geometry of molecules. This is a very valuable tool of MarvinSketch. And you see, structure 3 and structure 4 have different geometries, and the different configuration assignments are correct. The methylene bridge is in 3 above and in 4 below the drawing plane of the six membered ring (Of course the six membered rings are in reality not planar having boat conformation).

This corresponds to the IUPAC recommendations 2006 (Jonathan Brecher, Pure Appl. Chem. Vol. 78, 10, 1897-1970, 2006). The attached copy is from side 1916 and the 3D picture at right is produced using MarvinSketch and MarvinView.

The conclusion about structure 3 and 4 is: The methylene bridge and the bridgehead substituents must be on the same side of the ring drawing plane. And the molecule 3 and 4 have different geometries being diastereomers. So Marvin is interpreting these structures correctly.

And once more my conclusion: Only formulas using wedge bonds can be correctly interpreted by program algorithms, just as Marvin does it.

Best regards, Hans-Ulrich

Yes. I'm presuming that there are many other templates used by Marvin which do not have the problem we are discussing here, so I would be hesitant to turn template cleaning off entirely.

I'm presuming that there are many other templates

 used by Marvin which do not have the problem we are discussing here, so

 I would be hesitant to turn template cleaning off entirely.

If this is the case, from Marvin 5.11 you can solve it with 'Toff' option:

Toff

Turn off system template file definition usage during template based cleaning
(only local template file definition is used).

Solution:

1) Collect the template molecules you don't want to ignore (want to use) into the local template file definition .chemaxon/cleantemplates.t in the user's home dir.

By default marvin uses the following templates from which you can freely choose

chemaxon/marvin/templates/bicycles.t

cemaxon/marvin/templates/crown_ethers.t

chemaxon/marvin/templates/heterocycles.t

chemaxon/marvin/templates/cleantemplate.cxsmi

chemaxon/marvin/templates/fullerenes.t

chemaxon/marvin/templates/bridged_polycycles.t

chemaxon/marvin/templates/cleancage.cxsmi

2) Use 'Toff' cleaning option and only your selected templates will be used.

Hi Volfi,

sounds very good !

Questions: "Toff" will be a new option in version 5.11, yes ? Can we get a beta version to test it?

Best regards, Hans-Ulrich

With this functionality yes.

As soon as it will be available I will write to this topic also.

This sounds like a good solution, since your post indicates the list of available templates isn't long. It is unfortunate that new or updated templates will have to be manually checked for and added when upgrading ChemAxon, but otherwise I don't see drawbacks.

Krishna

We have not added new template file used by clean for a two years, just added templates to the already existing ones, so I dont think it is going to be a problem.