3d generation and oxy

Hi,





Using the clean 3d method (jchem v3.04), we have some trouble with conformers having an oxy function. For example, according to our chemists, atoms in functions such as (C=O)-O-X (ester, carbamate ...) the four atoms should be in the same plan. They are not in conformers generated. Is that normal ?





Thanks,





Vincent

vincentberenz wrote:

Hi, Using the clean 3d method (jchem v3.04), we have some trouble with conformers having an oxy function. For example, according to our chemists, atoms in functions such as (C=O)-O-X (ester, carbamate ...) the four atoms should be in the same plan. They are not in conformers generated. Is that normal ? Thanks, Vincent

Hi Vincent,





You and your chemists are right, however, it is not that easy to teach MM forcefields to such rules. The barrier is very low for that rotation anyway, so I think this difference is acceptable. I always say that if someone really wants accurate structures than optimize also the structures with some quantum chemical method. I suggest at least B3LYP/6-31G(d) level. I am calculating the barrier for the C-O rotation and then let you know.





Best wishes,





Ödön

vincentberenz wrote:

Hi, Using the clean 3d method (jchem v3.04), we have some trouble with conformers having an oxy function. For example, according to our chemists, atoms in functions such as (C=O)-O-X (ester, carbamate ...) the four atoms should be in the same plan. They are not in conformers generated. Is that normal ? Thanks, Vincent

Hi Vincent,





Thanks for this post, because it seems you've found a bug. I've tested both, the DFT and MM potential energy curves with Gaussian for HCO-OMe and scanned the O=C-O-C dihedral angle. See the attached file for the results. It's clear, that both MM and DFT have minimum at 180 degrees but only DFT has at 0 which is the lower energy minimum. The energy values has been adjusted to meet at 0 degrees for both method. This means that not having a planar structure in Dreiding is a bug for this structure, we will fix it ASAP.

Dear Odon,





Thank you for your answer.





"I always say that if someone really wants accurate structures than optimize also the structures with some quantum chemical method. I suggest at least B3LYP/6-31G(d) level"





Here is the code I use in java programs to retrieve conformers :





d2.clean(3,"S{conformers}");


CallbackIface mm = (CallbackIface)MarvinModule.load("Clean3D",null);


Object o = mm.callback("calcConformers",d2);


conf = (Molecule[])o;





I know there are options for 3D cleaning, but I did not find documentation in the api (the link in MoleculGraph, method clean, "cleaning options" does not seem to work). Where may I find some docs about cleaning option and the use of, for example, B3LYP/6-31G(d) level ?





Thanks,





Vincent

vincentberenz wrote:

Dear Odon, Thank you for your answer. "I always say that if someone really wants accurate structures than optimize also the structures with some quantum chemical method. I suggest at least B3LYP/6-31G(d) level" Here is the code I use in java programs to retrieve conformers : d2.clean(3,"S{conformers}"); CallbackIface mm = (CallbackIface)MarvinModule.load("Clean3D",null); Object o = mm.callback("calcConformers",d2); conf = (Molecule[])o; I know there are options for 3D cleaning, but I did not find documentation in the api (the link in MoleculGraph, method clean, "cleaning options" does not seem to work). Where may I find some docs about cleaning option and the use of, for example, B3LYP/6-31G(d) level ? Thanks, Vincent

Hi Vincent,





Sorry for the misleading answer. Quantum chemical (QC) calculations are not - yet - available in marvin. I also involved in the development of a quantum chemical program package, called Gaussian (www.gaussian.com), so I adviced this. Usually a QC calculation takes much-much more time than doing MM but also it is possible to get much more accurate answers. MM is usually good as a first guess and it has the advantage of being fast. I think that when the number of structures to study can be reduced (e.g. based on MM studies) then it is better to do some QC calculations at the end. We have plans to add more MM parametrization types (like AMBER, GROMOS) to Marvin, but MM will never give accurate structures for everything.





Best wishes,





Ödön

Dear Odon,





- Do you know approximativly when this bug (ester and amid not plan) will be resolved ? (point pretty critical for us)





- When will the MM parametrization (amber, Gromos) availble ?





Thanks,





Vincent

vincentberenz wrote:

Dear Odon, - Do you know approximativly when this bug (ester and amid not plan) will be resolved ? (point pretty critical for us) - When will the MM parametrization (amber, Gromos) availble ? Thanks, Vincent

Hi Vincent,





Sorry for the late answer, I had classes and had no chance to get e-mail...


We have found the source of the problem and can very likely make Dreiding to behave similarly like the QC method did. This change will fit in the spirit of Dreiding, so even without going for another forcefield we can solve your problem. It will be included in the next release of Marvin which might happen before the UGM.





Best wishes,





Ödön

Hi Odon,





Looking some conformers more closely, we found that, on the molecules we looked at, only two differents conformers were generated. For example, if ten conformers were generated, nine are almost superposable with only tiny differences, and only one is different from the nine others.





- Is this normal ?





- Is is possible to get documentation about how, using dreiding energy, you generate conformers ?





- Is there some option of the clean 3d we can use ?





Thanks,





Vincent

vincentberenz wrote:

Hi Odon, Looking some conformers more closely, we found that, on the molecules we looked at, only two differents conformers were generated. For example, if ten conformers were generated, nine are almost superposable with only tiny differences, and only one is different from the nine others. - Is this normal ? - Is is possible to get documentation about how, using dreiding energy, you generate conformers ? - Is there some option of the clean 3d we can use ? Thanks, Vincent

Hi Vincent,





We are working on this issue. currently there is no external option to change the behavior of the equivalent stucture filter. It is done automatically, but if you can send me some badly behaving example it could definitely help us and we can provide a better filter for the next release. I have one good news, the ester problem has been solved already (will be in the next release).





Best wishes,





Ödön

Dear Ödön,





Thank you for your answer. However, it does not help us totally. We are interesting in getting some information on the way you generate the different conformations. Which method are you using (systematic, Monte Carlo, ...)?


Apparently, there is no method to filter similar conformations. Am I right?


It is also strange that only two different conformations are generated even if several degree of rotation are present in the molecule (for example an alkyl chain with more than three carbons). Do you have explanation on this?


It is not clear for us in your answer if there are options for the Clean3D method. Could you give us more details on this?





Thank you in advance for your answers.





Frederique & Vincent

vincentberenz wrote:

Dear Ödön, Thank you for your answer. However, it does not help us totally. We are interesting in getting some information on the way you generate the different conformations. Which method are you using (systematic, Monte Carlo, ...)? Apparently, there is no method to filter similar conformations. Am I right? It is also strange that only two different conformations are generated even if several degree of rotation are present in the molecule (for example an alkyl chain with more than three carbons). Do you have explanation on this? It is not clear for us in your answer if there are options for the Clean3D method. Could you give us more details on this? Thank you in advance for your answers. Frederique & Vincent

Dear Frederique & Vincent,





There is a method in Clean3D for filter the similar conformers but at certain places it used too strict criteria and did not sign similarity.


The current pre-release version gives no more 3 different conformers for C4, 5 for C5 which are the correct numbers.


You can get the detailed list of options either by running





molconvert -H3D





or by using "h" or "?" in the option string, like:





d2.clean(3,"h");





I'll be glad to answer your specific question about the options.





Best wishes,





Ödön