Posted: Sun Mar 02, 2008 8:17 pmPost subject: Conformers and energies for chiral molecules

Hi,
now I am totally confused. Please look at the screenshot. I have done the same on the left and the right side. First I transferred the alanine template to the window. Then I used the Conformation - Conformers tool.
Left side: Maximum number of conformers. Right side: Calculate lowest energy conformer.

1) The energies of this conformers are totally different.
2) In the calculation result it is the other enantiomer.
Of course enantiomers have the same energy, but the calculation should not change the configuration.

What's wrong ?

Hans-Ulrich

Gabor ChemAxon personnel
Joined: 29 May 2005
Posts: 317

Posted: Mon Mar 03, 2008 12:52 pmPost subject: Bugs

Dear Hans-Ulrich,

Thank you for the bug report. You are absolutely right, the chirality should not be changed during the conformer generation. Energies also should be consistent and practically the same also.

We will track down these bugs (hopefully by the next release) and will keep you informed.

Posted: Fri Mar 21, 2008 12:11 pmPost subject: Three problems with energy calculations

Dear Gabor,
I have evaluated Marvin version 5.0.2. I have repeated the calculations for the simple molecule 2S-alanine and the results for "calculating all conformations" are numerically the same. But I got a new result for calculating the "lowest energy conformer", see screenshot 413 right hand side. The difference is about 5 kJ/mol.
Calculating the "geometry energy" gives the third value for the same molecule, see screenshot 412.

I think there a three problems discussed in this topic and perhaps it would be better to make three topics.

1) It is obviously difficult to find the conformer with the lowest energy.

2) There is a large difference between the table with all conformers and the single calculation for the "lowest energy conformer" (both in the tool conformers as in the tool geometry). The difference in the alanine example is between about -17 kJ/mol and 52 kJ/mol = about 69 kJ/mol.

3) The configuration is changing from the original molecule to the conformers from S to R in the conformer tool, but it is not changing in the geometry tool.

How to fix these three problems ?

Regards, Hans-Ulrich

Gabor ChemAxon personnel
Joined: 29 May 2005
Posts: 317

Posted: Fri Mar 21, 2008 6:54 pmPost subject: Bugs

Dear Hans-Ulrich,

We haven't fixed these issues yet, so the bugs are still present in the current version. You are right, the configuration should not be changed during the 3D clean, and it is also problematic that multiple "hyperfine" runs produce structures with significantly different energy.

I will post a note in this topic when these bugs are fixed. Tuning the actual clean3D settings to find the lowest energy conformer surely worth a further discussion, but this should be postponed until the above mentioned bugs are fixed.

Hi, I am having a similar problem with glycerol: the energie of lowest energie conformer is totally different to the energies of calculated conformers. I am not using enantiomers.
In fact, up to now this is the first time I saw these energy diferences with Marvin 5.0.2

Edmundo

Gabor ChemAxon personnel
Joined: 29 May 2005
Posts: 317

As far as I know we have not introduced any changes in 5.0.2 which could affect the artifacts detailed in this topic. If the issues above mentioned are fixed I will post a notification in this topic.

Hi Gabor,
one problem seems to be in the procedure to find the possible conformations and their energy minima. Please look at the calculation for cyclopropene, see screenshot458. There are two conformers ? Cyclopropene has no different conformers. The conformer No 1 has two hydrogens out of the plane of the three membered ring ?

Here the energies of conformer No 2 and the Geometry calculation are the same, both with the hydrogens at C1 and C2 in the plane, see screeshot459.

Regards, Hans-Ulrich

Gabor ChemAxon personnel
Joined: 29 May 2005
Posts: 317

Thanks for the screenshots. On each generated conformer a geometry optimization is invoked so the non-planar H atoms may result from an error in the force field (Dreiding) used. The other option is a potential flaw in the implementation of the handling of "hyperfine" finalization. The later seems more plausible from the discrepancies found around it. Btw, the planar structure definitely should have lower energy, so I will also check the FF.

Hi Hans-Ulrich, I am sending you the screenshot of calculated energies for glycerol conformers. Before all my calculations I always use a "clean 3D" of my molecules.

Posted: Sat Apr 12, 2008 5:24 pmPost subject: Glycerol example

Hi,
the problem is, I got other values for glycerol.
For "Calculate lowest energy conformer" the energy value was 3.94 kcal/mol.
Calculating all conformers showed 150 conformers for glycerol going from -26.42 to +4.04 kcal/mol.

Hi, further evaluating the conformers and energy problem, I found for the glycerol example the following.

1) After adding the hydrogens and "Clean 3D" there is the molecule in the MarvinSketch window as you see in the screenshot 466. Now calculating the Dreiding energy with the Never option I get 3.99 kcal/mol and it is the same conformation as in the MarvinSketch window. OK.

2) Calculating the Dreiding energy with the option "Calculate for lowest energy conformer Always" I get the same conformation and the energy is 3.94 kcal/mol, see screenshot 467. It seems to me the lowest energy conformer is not calculated. There are a lot of conformers with lower energy, see the two posts above.

There is one thing more: Calculating the "lowest energy conformer" with the "Geometry" tool takes about 2 seconds. But calculating all 150 conformers for glycerol takes about 175 seconds an my machine. Why is the "Geometry" tool so fast ? I think the "Geometry" tool must also calculate all conformers to know which one is the lowest energy conformer.

Regards, Hans-Ulrich

Gabor ChemAxon personnel
Joined: 29 May 2005
Posts: 317

The small inconsistency involving energies (3.99 vs 3.94 kcal/mol) calculated with/without invoking the coordinate generation process is already noticed in an other topic (http://www.chemaxon.com/forum/viewtopic.php?p=15990#15990).

The Geometry plugin is not intended to achieve a full (or even a deep) conformational analysis, since finding the global energy minimum necessarily requires an extensive sampling of the conformational space. (This process is covered by the Conformers tool.) It might be more straightforward to use "Invoke 3D coordinate generation" in the Geometry tool options dialog.

Hi Gabor,
I aggree, the method should be named "Invoke 3D coordinate generation". But there is still a problem: Which Clean 3D method to use ?

Please look at the glycerol in the conformation you can see in the screenshot. I have produced it using the MarvinSpace feature Miklos showed me in the topic ( http://www.chemaxon.com/forum/ftopic3732.html ) I transferred it to MarvinSketch and then immediately calculated the Dreiding energy using the option "Always". The conformation was changed and I got the conformer with 3.94 kcal/mol you can see in the post above. Using the option "Never" the conformation was not changed and I got the Dreiding energy you see in the screenshot below.

Obviously the Geometry option "Always" uses first the "Clean 3D Fine Build" method. I think it should use "Clean 3D Geometry Optimize" which does not change the conformation. I evaluated these effects using the Clean 3D methods alone on the Desk.

But one question more: What does the Geometry option "Never" do ? Does it any optimisation at all, bond lengths, bondangles or dihedral angles ? If not, the energy result is only an approximation.

Regards, Hans-Ulrich

Gabor ChemAxon personnel
Joined: 29 May 2005
Posts: 317

Since "always" was checked, a 3D coordinate generation process (to be more precise, the "Fine" process) was invoked on the structure regardless of its geometry. This process considers only the topology and the stereo informations only, not the actual conformation. It should not use geometry optimization alone, since a 2D (or even 0D if used from command line on SMILES input) coordinate should not be used as starting geometry for an optimization. (Note that the Fine process produces optimized coordinates.)

Option "Never" leaves the structure intact. You are right, the energy should be calculated on an optimized structure to have any meaningful result.

Currently we optimize the structure in the cartesian coordinate system using an own implementation of the Dreiding force field as the energy source. Since this force field contains energy components for bond lengths, angles and dihedrals, these components are also optimized. The optimization process is invoked on a starting geometry (generated either by Clean3D or given by the user) and is intended to find the closest energy minimum point. (False energy minimum found sometimes, hyperfine is intended to eliminate them.) Consequently, we currently do not optimize in internal coordinate system (where the structure is represented by bond lengths, angles and torsion values), neither we do global optimization (which finds the absolute energy minimum).

So, the result of a Clean3D call should return structures already around a local energy minimum (within a certain tolerance) since it invokes optimization at the last step. I would like to note that an inconsistency involving the results of hyperfine runs indicates that the results might not been optimized correctly; this bug will be investigated soon.

Dear Gabor,
one result of our discussion is: The text descriptions of the tool Geometry must be corrected. And also in the help text, see screenshot 471.

My idea is, to have the following options and descriptions:
1) For 2D structures: Build 3D and optimize geometry
2) For 3D structures: Find the closest energy minimum geometry for given structure

The other options are still applicable.

Best regards, Hans-Ulrich

Gabor ChemAxon personnel
Joined: 29 May 2005
Posts: 317

Posted: Thu Apr 17, 2008 3:58 pmPost subject: Thanks for the suggestions

Dear Hans-Ulrich,

Thank you for remark; the help clearly needs further clarification. I think your first suggestion is equivalent to the current "If molecule is in 2D" option. The second one (invoke optimization only) is definitely a reasonable choice to have.
I would go further and add an other option which invokes 3D clean (and optimize) for 0D or 2D structures and optimize only the 3D ones.

So, the Geometry plugin should have the following options:
-Dreiding optimization only when structure 3D, Clean3D if 2D
-Always do Clean3D
-Clean3D if structure is 2D (and leave 3D structures intact)
-Invoke Dreiding optimization only
-Do not modify structure (former "never" option)

The label for the choices in the plugin user interface should be "Invoke 3D coordinate calculation/optimization".

1) If structure is 0D or 2D
Clean 3D Hyperfine and invoke Dreiding optimization
2) If structure is already 3D
Do not modify geometry and calculate Dreiding Energy
Invoke Dreiding optimization only
Clean 3D Hyperfine and invoke Dreiding optimization

Do you see more options ?

Regards, Hans-Ulrich

PS: When can I get a prerelease to test it ;-)

Gabor ChemAxon personnel
Joined: 29 May 2005
Posts: 317

I think the former "never" option should be accessible as a fifth choice. The current "hyperfine" implementation can be extremely slow, so i think it must be an optional choice.

We are still working on the major modifications of Clean3D, so as soon as possible i will look for the possibility to pass it to you before its official release.

regards,
Gabor

Gabor ChemAxon personnel
Joined: 29 May 2005
Posts: 317

A couple of problems are mentioned in this topic. I would like to inform you that some of these are fixed in the forthcoming 5.1 version (its release is expected in a few days):
-A bug is fixed which sometimes corrupted the chirality of atoms during 3D clean
-As discussed in http://www.chemaxon.com/forum/viewtopic.php?p=16602#16602 hyperfine parameters are tuned to provide consistent results over consecutive runs (i would like to mention that further modifications are expected on this area in the near future)
-As discussed in http://www.chemaxon.com/forum/viewtopic.php?p=15990#15990 the energy calculation code is modified, now it gives consistent results. (I would like to note that new dreiding implementation is introduced, so its calculated energies might differ from the old one.)

The mentioned GUI modifications (affecting the Geometry plugin and the Conformers plugin "Calculate lowest energy conformer" option) are still in the queue.

this morning I presented shortly the Marvin Conformers calculations in our principal organic chemistry lecture. In the little discussion afterwards with the organic chemistry professor he asked immediately, "how to find the different conformers", knowing this is not an easy task !

One question: Does "new dreiding implementation is introduced" mean there is a new parameter set, or it is still based on "Goddard III"s paper ?

Regards, Hans-Ulrich

Gabor ChemAxon personnel
Joined: 29 May 2005
Posts: 317

The new implementation is still based on the referred paper.
Interpretation of bond angle and inversion terms are modified to ensure continuous second derivates around transition points to help optimization process.
Energy term for handling hydrogen bonds is currently disabled.
It includes some bugfixes, futher fine tuning (to handle non specified structures, for example organometallic complexes).

To watch this topic for replies
Register (enables digests) or give email address

You cannot post new topics in this forum You cannot reply to topics in this forum You cannot edit your posts in this forum You cannot delete your posts in this forum You cannot vote in polls in this forum You cannot attach files in this forum You cannot download files in this forum