resonance smiles

I'm trying to return Lewis resonance structures from a single window in an embedded MSketch applet, but all my smiles strings appear identical. I'm guessing there's canonicalization going on in the background.





On the Chemaxon-provided "Student Examination" page all the smiles are identical and I can't figure out how the lone pairs are mapped:





[O-][N+]([O-])=O.[O-][N+]([O-])=O.[O-][N+]([O-])=O |lp:0:3,2:3,3:2,4:3,6:3,7:2,8:3,10:3,11:2|





How can I tell whether a student submitted all correct resonance structures instead of three times the same thing or the same structure (bonds) but merely the lone pairs moved around?





Since I can't use the resonance arrow in this example, the applet returns a salt format (dot-separated) formula. What is its effect on the "lp" string? What if I used/required a resonance arrow on my server?





When I try to replicate this page on my server, I can only retrieve the smiles portion, but not the lp portion. From the source code (or documentation I could find) it's not at all obvious how to return both. What is the proper msketch_param or other variable that controls that?








This is all happening with MarvinSketch 5.1.3_2/build 2008-11-13.

Hi,





The smiles part are the same for all structure as smiles is not able to describe explicit lone pairs, but the extended part should be different.





The indexes of the atoms followed by a colon character and the number of


explicit lone electron pairs are written after "lp:".

Code:

When I try to replicate this page on my server, I can only retrieve the smiles portion, but not the lp portion. From the source code (or documentation I could find) it's not at all obvious how to return both. What is the proper msketch_param or other variable that controls that?

    This feature appears in Marvin 5.2 coming out soon. The cxsmiles will contain this information by default. The "lp:" portion will also be mentioned in the documentation.





Andras

Quote:

"The smiles part are the same for all structure as smiles is not able to describe explicit lone pairs, but the extended part should be different."

   





As shown in the image in my original post, not all smiles


should be the same. In resonance structures the atoms retain identity,


but the electrons are moved around. Since bonds represent electron


pairs it is important to know which atom pairs are single, double, or


even triple bonded. The three structures shown differ in this respect


and should therefore have different smiles strings. If only the lp


portion is distinct, how does one know whether the student actually


moved bonds around to obey the octet rule? The image attached here shows the exact same structure three times instead of proper resonance forms, yet it generates the same smiles + lp string! That is just bad science (and an opportunity for students to cheat the system).





As far as the version number: I'm using 5.1.3 and it is showing the lp on the demo page, but it won't give me the lp info on a page I coded myself using the page source code from the demo page. I'm not using a pre/beta-release at all.

Hi,





As far as I see  2009-01-15_154027.jpg shows three identical structures. Rotating the upper center molecule by 120 degrees counterclockwise you can get the left one, rotating further you can get the right one. You may have a misleading concept that smiles sting retrains atom indexes, but it is not true. You should make your atoms discrete to able to differentiate these structrures. I would suggest to use atom maps. So map your three oxigen with different values and you will get different smiles strings.





Andras

Quote:

Hi, As far as I see  2009-01-15_154027.jpg shows three identical structures. Rotating the upper center molecule by 120 degrees counterclockwise you can get the left one, rotating further you can get the right one.

   





Lewis structures are operationally invariant, which means that you cannot perform rotations or symmetry operations to get from one structure to another. Pretend the oxygen atoms are isotopically labeled. That would prohibit any possible transformation. The only things that move around in Lewis structures are electrons. You are right in the sense that nitrate is always nitrate, no matter which way you draw it, but the three Lewis resonance structures drawn are not identical at all.

Quote:

You may have a misleading concept that smiles sting retrains atom indexes, but it is not true.

   





There is no reason why, for instance, the =O in nitrate should always come last in a SMILES string. I have enough experience with SMILES and SLNs to know that I can write the smiles [O-][N+](=O)[O-] for nitrate instead of the [O-][N+]([O-])=O that MSketch prefers. In that sense, the SMILES string is indeed indexed. It's not until some sort of canonicalization is applied that all three resonance structures are turned into the same canonical SMILES.





Since bonds represent electron pairs and lone pairs represent electron pairs and radicals represent single electrons it is the sum total of combinations that needs to be evaluated to see whether all answers are indeed correct resonance structures. In canonicalizing the SMILES strings, yet leaving the lp string unaffected, the possibility of correctly evaluating the answer given is severely impaired: I may be able to evaluate a single Lewis structure, but I cannot distinguish between three valid resonance structures and three instances of the same Lewis structure.

Quote:

You should make your atoms discrete to able to differentiate these structrures. I would suggest to use atom maps. So map your three oxigen with different values and you will get different smiles strings.

   





Since the example given by ChemAxon was billed as "Student Exam" I would expect it to yield a reliable enough answer so I can disciminate between correct submissions by students and incorrect ones. In ChemAxon's example I cannot distinguish between the correct resonance structures submitted in my first image and the incorrect resonance structures submitted in my second image. Moreover, I can hardly expect students to map oxygen atoms through the GUI. I could present them with three separate nitrate skeletons to start with, but that would rather give away the answer to the question...

Quote:

Andras

Hi,

Quote:

Lewis structures are operationally invariant, which means that you cannot perform rotations or symmetry operations to get from one structure to another. Pretend the oxygen atoms are isotopically labeled. That would prohibit any possible transformation.

      





If the oxigen atoms are isotopically labeled then they are not symmetric structures, however in this case they are not labeled at all. You pointed out the problem, the labeling is missing. If you label your atoms some way the smiles will be different.





About your SMILES comment. You are right that one can use SMILES which is not canonical, but I don't know why would this be a solution for the problem. For example let the 1st 2nd and 3rd atomic indexes are the three Nitrogen atoms in the three separate structure, so you can get a SMILES something like: [N+]123.N456.N789.[O-]1.[O-]2.O=3.O4.O5.O6.O7.O8.O9





But I still don't know how can you compare two SMILES string which is exactly the same structure (e.g. center structure in 2009-01-15_154027.jpg) but the atoms are indexed in different ways. It would generate three different SMILES string.

Quote:

In ChemAxon's example I cannot distinguish between the correct resonance structures submitted in my first image and the incorrect resonance structures submitted in my second image.

      


Without correspondence of the atoms the correct and incorrect images are just the same for me. Please explain why the incorrect answer is incorrect if the Oxigen atom nuclei are not distinguishable.





I think if I can properly understand the problem why the incorrect and the correct answer is different then I can find the solution.





Andras

Hi Andras,


This is pretty basic chemistry that we teach in our introductory chemistry classes.


I can't give you a copy of our textbook but the Wikipedia entries might help: http://en.wikipedia.org/wiki/Lewis_structures and http://en.wikipedia.org/wiki/Resonance_structure


It seems that you are missing one of the key concepts here: the structures are operationally invariant, or in other words: you are not allowed to perform any transformation (including rotation) to "convert" one structure into any other structure. The only particles that are allowed to move around in the image are subatomic particles, namely electrons.


Since each bond represents an electron pair that is mutually shared between two adjacent atoms, it actually matters which two atoms are involved. The double bond in the first image is sometimes pointed down, sometimes pointed up/left, and sometimes pointed up/right. The atoms are not interconvertible. What happens is that two pairs of electrons are being pushed around. First one of the three pairs from one of the single-bonded O is pushed into the bond to the N making for a double bond, and next one pair of electrons is pushed out of the existing double bond into the previously double-bonded O. The atoms do not move at all in the process.


Each bond is a pair of electrons, as is each "lone pair". Each radical is a single electron, but we are not concerned with that just yet. If we let ":" represent an electron-pair, I can rewrite   :::O-N=O:: <-> ::O=N-O:::   to   :::O:N::O:: <-> ::O::N:O:::   . Since I am not allowed to "flip the molecule around", these two (partial) resonance structures are distinct. In fact, they represent two extreme situations that describe some of the properties of the "average" molecule. To quote from Wikipedia: " The individual resonance structures do not exist in reality: the molecule does not inter-convert between them."


Thus, if I have a SMILES [O-]1[N+]2([O-]3)=O4 it is distinct from the SMILES [O-]1[N+]2(=O3)[O-]4. But that is not the whole story: I also have to count the lone pairs on every single atom to make sure they follow the "octet rule": http://en.wikipedia.org/wiki/Octet_rule


Whereas in the first image I submitted, electrons are being moved around to yield three distinct resonance structures, in the second image nothing is being moved around. In the first case, the student knows the significance of all the Lewis rules, in the second case the student is either trying to cheat the system or doesn't know what is expected. In both cases, though, MSketch generates the exact same SMILES+lp answer key, which means that I cannot distinguish the good students from the bad (and makes the application useless for this purpose).


Now, if MSketch did not automatically rewrite all SMILES strings into a canonical form (putting the =O always last) I could write an algorithm to determine whether the total number of electrons in bonds + lone pairs + radicals is correct or not. What I cannot do is expect the students to label all atoms consistently through the GUI, nor can I give them labeled template SMILES to begin with because it would give away the expectations for the answer (for instance how many resonance structures are expected).


I hope that clears up some of the misunderstanding of what I am trying to achieve here.

Maybe you are thrown off by the obvious symmetry in the molecule...





Try it with this one:





[O-][P+](Cl)(Cl)Cl.O=P(Cl)(Cl)Cl

These are not resonance structures:








F[Xe](F)(Cl)Cl.F[Xe](Cl)(F)Cl





But MSketch generates the following SMILES+lp string anyway:





F[Xe](F)(Cl)Cl.F[Xe](F)(Cl)Cl |lp:0:3,1:2,2:3,3:3,4:3,5:3,6:2,7:3,8:3,9:3|





There even isn't a way to turn one into the other without breaking some bonds and moving atoms around. So why does MSketch generate an identical SMILES for each molecule?

Thank you for the detailed answer.





I think it is much clearer now. I have some technical question.


I assume that the student would see an msketch applet to draw his answer.


How can you guarantee that the student will use the same atom indexes for the three correct answer what you would use and not take the first molecule and rotate it or draw his structure starting form a different atoms?





Taking your example I have drawn the molecule starting from different atoms (see attached picture lone electron pairs are not shown) in this case the non canonical SMILES would be:








[O-][N+]([O-])=O.[O-][N+]([O-])=O.[O-][N+]([O-])=O





But this is exactly the same result as the student would just copy the first molecule two times. So how can you differentiate the two method in this case.





Andras

Quote:

These are not resonance structures: F[Xe](F)(Cl)Cl.F[Xe](Cl)(F)Cl

  





This two structure is chemically the same.

Quote:

There even isn't a way to turn one into the other without breaking some bonds and moving atoms around. So why does MSketch generate an identical SMILES for each molecule?

 


I don't really understand why someone would break bonds or move atoms around if it is the same chemical structure.





Probably we have different interpretation of SMILES string so let me write my interpretation:





F[Xe](F)(Cl)Cl is a molecule which have a Xe atom connected to two F atoms and two Cl atoms. (The chemistry does not know about atom indexes it is just necessary for computer representation.)





Please let me know your interpretation.

Quote:

Quote: These are not resonance structures: F[Xe](F)(Cl)Cl.F[Xe](Cl)(F)Cl      Quote: This two structure is chemically the same.   These two structures are not the same. One is a cis-, the other a trans-geometric isomer. This is part of the reason why we need to talk about Lewis structures in introductory chemistry texts. The lone pair electrons keep the four ligands in a plane. Therefore there is no possible rotation that will convert one into the other, and the compounds are therefore distinct. Quote: There even isn't a way to turn one into the other without breaking some bonds and moving atoms around. So why does MSketch generate an identical SMILES for each molecule?     Quote: I don't really understand why someone would break bonds or move atoms around if it is the same chemical structure. Probably we have different interpretation of SMILES string so let me write my interpretation: F[Xe](F)(Cl)Cl is a molecule which have a Xe atom connected to two F atoms and two Cl atoms. (The chemistry does not know about atom indexes it is just necessary for computer representation.) Please let me know your interpretation.

Quote:

Thank you for the detailed answer. I think it is much clearer now. I have some technical question. I assume that the student would see an msketch applet to draw his answer. How can you guarantee that the student will use the same atom indexes for the three correct answer what you would use and not take the first molecule and rotate it or draw his structure starting form a different atoms? Taking your example I have drawn the molecule starting from different atoms (see attached picture lone electron pairs are not shown) in this case the non canonical SMILES would be: C(=O)(O)O.C(=O)(O)O.C(=O)(O)O But this is exactly the same result as the student would just copy the first molecule two times. So how can you differentiate the two method in this case. Andras

 





 This is exactly what I am trying to figure out. The ChemAxon example that comes with the installation is billed as a "Student Exam" (http://www.chemaxon.com/marvin/examples/applets/sketch/studentexam/index.html), but how can I grade the answer correctly with what MSketch reports as the answer?





I like using MSketch for OrgChem problems and have been vexed by Lewis structure problems for a while. When I saw the Student Exam example in the new release I was happy to see that someone attempted to solve this problem. I can see this approach working for single Lewis structures, but I can't quite figure out how this would apply to resonance structures, which is what prompted my original post.

Quote:

These are not resonance structures: F[Xe](F)(Cl)Cl.F[Xe](Cl)(F)Cl These two structures are not the same. One is a cis-, the other a trans-geometric isomer.

  


I think you are not right in this question. Could you explain me my why do you think they are different geometric isomers?





I can agree that they are geometric isomers if square-planar parity is defined: F[Xe@SP1](F)(Cl)Cl but otherwise they are exactly the same molecule.





Please let me know what do you think.





Andras

Quote:

I like using MSketch for OrgChem problems and have been vexed by Lewis structure problems for a while. When I saw the Student Exam example in the new release I was happy to see that someone attempted to solve this problem. I can see this approach working for single Lewis structures, but I can't quite figure out how this would apply to resonance structures, which is what prompted my original post.

  





As you pointed out it is working for single Lewis structures. To work with resonance structures you should give the mapping of atoms to identify them.


I would imagine that you show an msketch applet loaded with one mapped molecule. And the task is:


Copy the original (mapped) molecule and rearrange electrons on the copy to depict resonance forms. Do not change the mapping of the atoms just lone electron pairs and bonds.





Or probably it is better to use an mview applet with the mapped original molecule in question (which cannot be changed) and an msketch applet


for the answer. So the student should copy from the mview applet to the msketch.





What do you think?





Andras

Quote:

These are not resonance structures: F[Xe](F)(Cl)Cl.F[Xe](Cl)(F)Cl These two structures are not the same. One is a cis-, the other a trans-geometric isomer.

            

Quote:

I think you are not right in this question. Could you explain me my why do you think they are different geometric isomers?

  





I think we are digressing somewhat from the original topic, but I think it is relevant anyway.





The student will probably be presented with a question along the lines of: "The compound XeF2Cl2 may exist in multiple geometric isomeric forms. Use Lewis structures to show and explain how this is possible." Or some variant thereof. In that case the student is expected to give the cis- and the trans-isomer showing the lone pairs on the central Xe atom. They are not resonance structures because you cannot interconvert the graphs by pushing electrons around.





If you plug the SMILES I originally posted (   F[Xe](F)(Cl)Cl.F[Xe](Cl)(F)Cl   ) in MSketch you get the graphs in the attached image. If you plug the canonicalized version of the SMILES returned by MSketch (   F[Xe](F)(Cl)Cl.F[Xe](F)(Cl)Cl |lp:0:3,1:2,2:3,3:3,4:3,5:3,6:2,7:3,8:3,9:3|   ) back into MSketch you get two indistinguishable graphs.





The real puzzle the student is asked to solve is this: why does the attached image represent two molecules, whereas the corresponding carbon-based compound does not represent two molecules?





The basic problem of presenting two molecules in the same MSketch windows remains. If you didn't quite grasp the significance when I first presented these isomers, how is a student supposed to enter the information correctly and how can I grade it correctly?





From the discussion so far, you seem to be steering me towards giving the student more information than I really want to. The student needs to do the analytical, integrative, and constructive work that leads to the correct answer.





I know what SMILES I could use to send the complete information to the student, but how can I retrieve an answer from the student that is intelligible and can be graded?





If there is another way to address this type of problem than extended SMILES, I'd love to hear about it, but it has to be applicable across the spectrum of question types, not just a single Lewis structure. Who at ChemAxon put the example page together and why did they think it was a good idea to use extended SMILES?

Quote:

I can agree that they are geometric isomers if square-planar parity is defined: F[Xe@SP1](F)(Cl)Cl but otherwise they are exactly the same molecule. Please let me know what do you think. Andras

Hi Michiel,

Quote:

If you plug the SMILES I originally posted (   F[Xe](F)(Cl)Cl.F[Xe](Cl)(F)Cl   ) in MSketch you get the graphs in the attached image. If you plug the canonicalized version of the SMILES returned by MSketch (   F[Xe](F)(Cl)Cl.F[Xe](F)(Cl)Cl |lp:0:3,1:2,2:3,3:3,4:3,5:3,6:2,7:3,8:3,9:3|   ) back into MSketch you get two indistinguishable graphs.

 


If you need coordinates loaded together with the structures, then instead of SMILES you should plug in a format that stores the coordinates, e.g. MRV.

Quote:

I know what SMILES I could use to send the complete information to the student, but how can I retrieve an answer from the student that is intelligible and can be graded?

 


I think you want to use an inappropriate format (i.e. SMILES) for your purpose. SMILES has its limitations; one of these is that the order of the atoms - if they represent the same connection - doesn't mean anything (without additional signs). 

Quote:

If there is another way to address this type of problem than extended SMILES, I'd love to hear about it, but it has to be applicable across the spectrum of question types, not just a single Lewis structure. Who at ChemAxon put the example page together and why did they think it was a good idea to use extended SMILES?

 


The Student Examination example page was put together to show the result of a custom development, where the purpose was to implement features that enables the examination of students on the subject of proper understanding of lone pairs and radicals. For that purpose, exporting the answer to canonical smiles  extended with the info of the lone pairs and radicals was a proper solution. Actually, we have a feature request from other clients to be able to also store the coordinates in ChemAxon Extended SMILES, but had no time to implement it yet. Would this feature be satisfactory for you?


Best regards,


Akos

Quote:

Hi Michiel, Quote: If you plug the SMILES I originally posted (   F[Xe](F)(Cl)Cl.F[Xe](Cl)(F)Cl   ) in MSketch you get the graphs in the attached image. If you plug the canonicalized version of the SMILES returned by MSketch (   F[Xe](F)(Cl)Cl.F[Xe](F)(Cl)Cl |lp:0:3,1:2,2:3,3:3,4:3,5:3,6:2,7:3,8:3,9:3|   ) back into MSketch you get two indistinguishable graphs.      Quote: If you need coordinates loaded together with the structures, then instead of SMILES you should plug in a format that stores the coordinates, e.g. MRV.   Quote: I know what SMILES I could use to send the complete information to the student, but how can I retrieve an answer from the student that is intelligible and can be graded?      Quote: I think you want to use an inappropriate format (i.e. SMILES) for your purpose. SMILES has its limitations; one of these is that the order of the atoms - if they represent the same connection - doesn't mean anything (without additional signs).    The more I get involved with this topic, the more I am convinced that order is everything and coordinates are really only substitutes for order. I played around with MRV files with the coordinate specifications removed and I got a lot closer to what I was looking for. The issue I can't solve from there is that our faculty need to be able to write their own questions with minimal knowledge of JScript or file formats. Whereas it is easy to pass a single-line variable to an application it gets a lot more complicated when (compressed) multi-line data need to be passed (and eventually parsed for proper grading). (I also can't get more than one resonance arrow saved, among other issues with that approach.) I had even written up a specification for a Lewis-structure tool before I found the Student Exam page. I think there is a fundamental difference in approach between general organic chemistry and introductory chemistry, specifically Lewis structures, in their assessment of what is important. For Lewis-structures, order is important since the molecules are operationally-invariant (all symmetry operations are prohibited), whereas for other types of chemistry Brownian motion is assumed and orientation is random. Yet, somehow, when I give two visually distinct SMILES (different order of elements given) I get two visually distinct graphs (e.g., the two geometric isomers of XeF2Br2). So, at some level, MSketch does make a distinction based on order, even though order is irrelevant per your assertion. If that order is significant when giving information to MSketch, why can't we give it the same significance when obtaining information from MSketch? If the order is significant, unwittingly or not, on the input-side, why can't we make it explicit on the output side? There has to be some sort of algorithm already in place that takes GUI coordinates and translates that into cxsmiles. Is it possible to add a switch (or modify the L switch) so we see a correspondence between order and lone-pair assignment? When I have three resonance structures, such as is the case with nitrate, I need to check that all three structures have a different bond-order, and that each atom has an appropriate number of lone-pairs given the bond order. The order in which the resonance structures are presented, i.e. their location on the page (read "coordinates"), is irrelevant. I don't really see how providing the coordinates would solve the problem of distinguishing the correct three resonance structures from the attempt to fool the instructor (see my original images of the nitrate graphs). Since I don't know where the request for coordinates originated, I don't want to disagree with the other users, but I'm not convinced it would address the problem I'm trying to solve. Quote: If there is another way to address this type of problem than extended SMILES, I'd love to hear about it, but it has to be applicable across the spectrum of question types, not just a single Lewis structure. Who at ChemAxon put the example page together and why did they think it was a good idea to use extended SMILES?      Quote: The Student Examination example page was put together to show the result of a custom development, where the purpose was to implement features that enables the examination of students on the subject of proper understanding of lone pairs and radicals. For that purpose, exporting the answer to canonical smiles  extended with the info of the lone pairs and radicals was a proper solution. Actually, we have a feature request from other clients to be able to also store the coordinates in ChemAxon Extended SMILES, but had no time to implement it yet. Would this feature be satisfactory for you? Best regards, Akos

Is it at all possible to return generic SMILES specifying some command option?





I don't think we've gotten to the bottom of this yet. I'm not sure generic SMILES would give us the solution to the problem, but the more I play around with MRV files, the less convinced I am that that would provide a solution.





If order is everything, then there has to be an algorithmic solution that is linked to the way structures are displayed and read from the GUI before it even gets translated into any other format...

Quote:

Is it at all possible to return generic SMILES specifying some command option?

Yes, but it is just less canonicalised SMILES. We always use some kind of canonicalisation.

Quote:

I don't think we've gotten to the bottom of this yet. I'm not sure generic SMILES would give us the solution to the problem, but the more I play around with MRV files, the less convinced I am that that would provide a solution

As far as I see using generic SMILES would not solve the problem.

Quote:

If order is everything, then there has to be an algorithmic solution that is linked to the way structures are displayed and read from the GUI before it even gets translated into any other format..

Coordinates are not taken into accout when SMILES format is generated (as SMILES strings have no coordinates).





But as far as I know the solution with the atom maps would be acceptable for you.








Andras

I just wanted to follow-up on this thread instead of leaving it as unfinished business.





After some phone conversations we agreed that it would be possible to use the atom mapping feature in combination with cxsmiles to achieve the desired effect. Since the atom mapping stays constant the order of the atoms in the smiles becomes irrelevant and we can then match the lone pair designations from the extended smiles directly to mapped atoms. (correct me if I'm wrong)





I haven't tested this solution with students yet, but my initial attempts appear to confirm this solution.

Quote:

Quote: Is it at all possible to return generic SMILES specifying some command option? Yes, but it is just less canonicalised SMILES. We always use some kind of canonicalisation. Quote: I don't think we've gotten to the bottom of this yet. I'm not sure generic SMILES would give us the solution to the problem, but the more I play around with MRV files, the less convinced I am that that would provide a solution As far as I see using generic SMILES would not solve the problem. Quote: If order is everything, then there has to be an algorithmic solution that is linked to the way structures are displayed and read from the GUI before it even gets translated into any other format.. Coordinates are not taken into accout when SMILES format is generated (as SMILES strings have no coordinates). But as far as I know the solution with the atom maps would be acceptable for you. Andras

.