Previous Thread
Next Thread
Print Thread
Page 2 of 2 1 2
Re: charge optimization for a sulfoximine molecule
biochemist #30621 09/06/12 10:17 PM
Joined: Nov 2010
Posts: 25
B
Forum Member
OP Offline
Forum Member
B
Joined: Nov 2010
Posts: 25
I did frequency calculation on the conformation I am using and there were no imaginary frequencies.

I tried to optimize the charges with all of these constraints on the carbons but it t is impossible to get charges optimized(msxn2.str, msxn2.ene). I calculated the MP2 Merz-Kollman charges to take a look at the QM charge distribution(picture attached). I kept the charges of c4, c19 as you suggested but changed the charges of c9 and c13 so that C9H3 and C13H2 groups have a net 0 charge as in the QM charge distribution and re-optimizing the charges, optimizing bond lengths and angles equilibrium values, reoptimizing the charges, I got better agreement between QM and MM but still not within convergence criteria(msxn6.str, msxn6.ene). Should I go on and optimize the frequencies?

Attached Files
msxn2.ene.txt (753 Bytes, 341 downloads)
msxn2.str.txt (4.21 KB, 339 downloads)
msxn6.ene.txt (756 Bytes, 367 downloads)
msxn6.str.txt (4.21 KB, 337 downloads)
msxn-chargedistrin-numbs.jpg (202.49 KB, 314 downloads)
Re: charge optimization for a sulfoximine molecule
biochemist #30623 09/07/12 12:48 AM
Joined: Dec 2005
Posts: 1,535
Forum Member
Offline
Forum Member
Joined: Dec 2005
Posts: 1,535
Originally Posted By: biochemist
I did frequency calculation on the conformation I am using and there were no imaginary frequencies.
Good for you, but I'm talking about the possibility that you might have missed a lower-energy gauche conformation because you started an optimization form an adjacent cis conformation that is higher in energy and the minimizer stayed stuck in said cis conformation, so then that cis conformation would have imaginary frequencies. It doesn't necessarily have to be the case, but it's something you might want to look into.

Originally Posted By: biochemist
I tried to optimize the charges with all of these constraints on the carbons but it t is impossible to get charges optimized(msxn2.str, msxn2.ene).
  • Just checking... O7 and N8 are not actually "out of the plane" as the text in the ene file suggests, are they?
  • The QM for O3 is different than in the last table. Is this a lone pair orientation? If yes, be advised that we usually consider both the trigonal planar and the lone pair conformation. We either target whichever's stronger, or go for a compromise between the two. So you have some leeway there. Also, can I see that lone pair conformation? Perhaps it's still not ideal; those things are tricky...

Originally Posted By: biochemist
I calculated the MP2 Merz-Kollman charges to take a look at the QM charge distribution(picture attached).
I'm not sure the Merz-Kollman can be trusted on a charged system. Especially if it doesn't have very small spatial extents and/or if it has atoms that are somewhat buried.

Originally Posted By: biochemist
I kept the charges of c4, c19 as you suggested but changed the charges of c9 and c13 so that C9H3 and C13H2 groups have a net 0 charge as in the QM charge distribution
Athough the resulting charge set is not that unreasonable, I'd like to repeat on principle that I don't think the QM charge distribution can be trusted. Also, you probably did this to channel more negative charge away from the central electronegative atoms (in comparison with MESN) in an attempt to make the MM interactions less favorable. I think this goal can be accomplished more effectively by moving negative charge from the nitrogen and oxygens into the sulfur and phosphorus. There's still plenty of room for doing so...

Also, just for the record, C9 is not exactly 0.09 more positive than C13 (it's 0.10) and C4 is -0.19 instead of -0.17 (but that's obviously nitpicking).

Originally Posted By: biochemist
I got better agreement between QM and MM but still not within convergence criteria(msxn6.str, msxn6.ene). Should I go on and optimize the frequencies?
I think the agreement for the charges is way to bad. Now that I looked at this new data, I'm starting to think that your QM water interaction energies are unusually weak for a negatively charged molecule across the board. Can you post your QM inputs and/or outputs? Or at least, a few of them? (say, O2, O7 and N8) If the QM target data somehow happens to be flawed, then we can spend a lot of time and never find an acceptable charge set.

Re: charge optimization for a sulfoximine molecule
Kenno #30626 09/07/12 04:01 PM
Joined: Nov 2010
Posts: 25
B
Forum Member
OP Offline
Forum Member
B
Joined: Nov 2010
Posts: 25
Originally Posted By: Kenno
Just checking... O7 and N8 are not actually "out of the plane" as the text in the ene file suggests, are they?
No. They aren't. I am sorry. I just forgot to delete these comments when I edited the input file.
Originally Posted By: Kenno
The QM for O3 is different than in the last table. Is this a lone pair orientation? If yes, be advised that we usually consider both the trigonal planar and the lone pair conformation. We either target which ever's stronger, or go for a compromise between the two. So you have some leeway there. Also, can I see that lone pair conformation? Perhaps it's still not ideal; those things are tricky
The QM for O3 in the last table is the trigonal planar with the water hydrogen oriented away from the ethyl group. I should have replaced the old QM O3 in charmm input with the O3 charge in the last table. I corrected that and attached the corrected msxn2.ene and msxn6.ene, however, the gab between QM and MM O3 is still very big. I also attached the picture of the lone pair conformation(The interaction energy is -3.87 Kcal/mol)
Originally Posted By: Kenno
Also, you probably did this to channel more negative charge away from the central electronegative atoms (in comparison with MESN) in an attempt to make the MM interactions less favorable. I think this goal can be accomplished more effectively by moving negative charge from the nitrogen and oxygens into the sulfur and phosphorus. There's still plenty of room for doing so...

Yes, Making the MM interactions less favorable is exactly what I want to do. Moving negative charge to sulfur and phosphorus didn't help as in (msxn2.ene).
Originally Posted By: Kenno
I think the agreement for the charges is way to bad. Now that I looked at this new data, I'm starting to think that your QM water interaction energies are unusually weak for a negatively charged molecule across the board. Can you post your QM inputs and/or outputs? Or at least, a few of them? (say, O2, O7 and N8) If the QM target data somehow happens to be flawed, then we can spend a lot of time and never find an acceptable charge set.

Sure. I attached the QM input and output files (and also MM input and output files just in case you want to look at them).

Attached Files
sxn-O3--HOH-lonepair.jpg (44.13 KB, 338 downloads)
posting.tar.gz (432.59 KB, 252 downloads)
msxn2.ene.txt (711 Bytes, 342 downloads)
msxn6.ene.txt (714 Bytes, 331 downloads)
Re: charge optimization for a sulfoximine molecule
biochemist #30688 09/13/12 07:00 PM
Joined: Nov 2010
Posts: 25
B
Forum Member
OP Offline
Forum Member
B
Joined: Nov 2010
Posts: 25
You are right. Th QM interaction energies are very weak for a negatively charged molecule. The main mistake was not in the input files or the water models but in calculating the QM interaction energies.
The QM interaction energy =
Energy(complex) - Energy(molecule)- Energy(water)

I was using the energy value of the HF optimized structure of molecule (already mp2 optimized), however, when I used the HF energy value (single point energy calculation)of the mp2 optimized molecule, the resulting energy value was about 6Kcal/mol lower and the QM Interaction energies values were much more stronger. The interaction energies after charge optimization are attached.

Attached Files
interaction-energies.JPG (55.51 KB, 286 downloads)
msxn2.str.txt (4.06 KB, 337 downloads)
Re: charge optimization for a sulfoximine molecule
biochemist #30691 09/14/12 12:55 AM
Joined: Dec 2005
Posts: 1,535
Forum Member
Offline
Forum Member
Joined: Dec 2005
Posts: 1,535
Glad you found your problem yourself; I've been absolutely swamped the last week. That's a dramatic improvement, isn't it?

For the record, we calculate the HF/6-31G* water interactions with the molecule of interest frozen at the MP2 optimized geometry and the water frozen at TIP3P geometry; only the interaction distance is allowed to vary. That way, we only have one E(molecule) (the HF/6-31G* energy of the MP2 optimized geometry) and one E(water) (the HF/6-31G* energy of water at TIP3P geometry) to subtract from each of the E(inter). It wasn't entirely clear to me whether this is what you did or not.

I'll have a closer look somewhere next week. As I said, it would be interesting to try to follow the guidelines in one of my previous posts (if you haven't done so already).

Re: charge optimization for a sulfoximine molecule
biochemist #30693 09/14/12 09:11 PM
Joined: Nov 2010
Posts: 25
B
Forum Member
OP Offline
Forum Member
B
Joined: Nov 2010
Posts: 25
Of course, The results are way better now.

I followed your previous guidelines and I am attaching a better optimized charges.

I am, however, worried about the charge of phosphorus. I noticed that all phosphorus(PG1) has a charge of >= +1, while the phosphorus I am optimizing is +0.22 . Is this OK?

I will go ahead and optimize the intramolecular parameters, I can always go back and reoptimize the charges if it still needs improvement.

Attached Files
msxn3.str.txt (4.06 KB, 343 downloads)
Interactionenergies.png (19.87 KB, 313 downloads)
Re: charge optimization for a sulfoximine molecule
biochemist #31100 12/12/12 05:13 AM
Joined: Dec 2005
Posts: 1,535
Forum Member
Offline
Forum Member
Joined: Dec 2005
Posts: 1,535
Sorry for my very late reply. We see dramatic shifts in charges like that relatively often when parameterizing functional groups that have a net charge. In general, I tend to worry less about partial charges becoming to small than about them becoming to big. In other words, the phosphorus charges are probably OK.

Re: charge optimization for a sulfoximine molecule
biochemist #31205 01/10/13 04:12 PM
Joined: Nov 2010
Posts: 25
B
Forum Member
OP Offline
Forum Member
B
Joined: Nov 2010
Posts: 25
I finished the parameter optimization for that sulfoximine compound. Thank you for all your help. However, I need to add amino group to the terminal methyl C19(amsxn.jpg). So, first I deleted one terminal hydrogen from C19-methyl group. Second, attached amino group(NH3+) to C19. My problem is C19 atom type changed now from CG331 in the previous optimization to CG324 after adding the amino group. The charge of C19 now should be 0.12 instead of -0.27 and I am afraid that this big shift in charge, in addition to the change in C19 atom type will affect all the parameters I have optimized. As a result, Is it necessary to repeat the whole optimization process for the whole molecule or is it enough to optimize the new dihedral parameters only?

Attached Files
amsxn.jpg (356.69 KB, 288 downloads)
Re: charge optimization for a sulfoximine molecule
biochemist #32578 09/09/13 07:40 PM
Joined: Dec 2005
Posts: 1,535
Forum Member
Offline
Forum Member
Joined: Dec 2005
Posts: 1,535
I know it's probably way to late, but for whoever stumbles upon this thread, the change in atom type should produce enough freedom in parameter space to get a good agreement by just optimizing the new dihedral parameters. Re-optimizing the other ones would arguably be undesirable because the resulting parameters would be very specialized and nontransferable. This is one of the very reasons why we have the divide-and-conquer strategy.

Re: charge optimization for a sulfoximine molecule
biochemist #32615 09/18/13 02:17 PM
Joined: Nov 2010
Posts: 25
B
Forum Member
OP Offline
Forum Member
B
Joined: Nov 2010
Posts: 25
That is what I did.

We developed new parameters for the N-phosphonosulfonimidoyl and acylphosphate functional groups following the standard protocol of CGenFF.The manuscript with all information regarding those parameters is accepted in Journal of Molecular Modeling and will be online soon. I will post the link later here in the forum for anyone who needs them.

Page 2 of 2 1 2

Moderated by  alex, lennart, rmv 

Link Copied to Clipboard
Powered by UBB.threads™ PHP Forum Software 7.7.4
(Release build 20200307)
Responsive Width:

PHP: 5.6.33-0+deb8u1 Page Time: 0.014s Queries: 40 (0.006s) Memory: 0.9966 MB (Peak: 1.1331 MB) Data Comp: Off Server Time: 2020-10-24 16:51:03 UTC
Valid HTML 5 and Valid CSS