VMD-L Mailing List

From: Vermaas, Joshua (Joshua.Vermaas_at_nrel.gov)
Date: Wed Sep 25 2019 - 09:38:35 CDT

Hi Per,

If you are happy with the charges on some atoms, just do not include them in the optimization. If memory serves, FFTK autopopulates the charge optimization table with the initial charges from the psf. If these entries are deleted, FFTK treats the psf charges for the deleted atoms as gospel and will fix those charges during optimization. During the optimization, the charges on the remaining atoms will be optimized so that the total charge for the whole molecule ends up being whatever you set.

My recommendation is that you tackle the net neutral molecule first. From what I know about ionization, usually there is substantial charge rearrangement to deal with the hydrogen that is lost. This means that the usual rules for CHARMM-based parameters (0.09 for most hydrogens, 0.42 or so if they are on an alcohol, etc.) do not really apply anymore, and you don't really know what kind of crazy answers you are going to get, which means that CGenFF's penalties may be inaccurate. Even worse, you'd need to run a whole different set of Gaussian calculations, so you may as well consider them to be two entirely different molecules.

A similar story applies to the dihedrals, in that FFTK can be told to only tune specific parameters by deleting the other parameters from the little table in the optimization setup.

-Josh

On 2019-09-25 01:21:20-06:00 Per Larsson wrote:

Hi JC,
Thanks for your input. This is also a good suggestion, and something that I also actually did. The resulting parameterization generally has very low penalties (zero in many cases, especially the fatty acid moiety). In fact, there is only one dihedral with a penalty score above 10, and three or four charges with penalties around 15.
But I am not sure how to proceed from here with ffTK. I do find the ”start from cgenff” tab, and can read in my resulting str-file. If you don’t mind me asking a few basic methodological questions then. For optimization of charges, do I only calculate water interactions for those atoms with a high charge penalty? Charge groups or not, the total charge on my molecule is either -1 (as the ionized species) or 0, so I need the sum of charges to sum to that. Do I include all the low-penalty charges as constraints somehow? Similar question also for the dihedral, do I specifically only scan that dihedral and optimize against QM-data?
Thanks for your patience
/Per

23 sep. 2019 kl. 00:04 skrev Gumbart, JC <gumbart_at_physics.gatech.edu<mailto:gumbart_at_physics.gatech.edu>>:

Good advice from Josh, as always. We go a step further by first running the molecule through CGenFF: https://cgenff.umaryland.edu/
and only then parametrizing those regions with high penalties. No need to reinvent the wheel for common moieties already present in the FF.
Best,
JC

On Sep 20, 2019, at 5:31 AM, Per Larsson <larsson.r.per_at_gmail.com<mailto:larsson.r.per_at_gmail.com>> wrote:

Hi Josh,
Many thanks for your reply. All of what you are saying makes sense. I will start by adding a shorter fatty acid chain as you suggest to the dihedral incorporated. Let me work on this some more and I’ll report back any additional problems (or success).
Thanks
/Per

19 sep. 2019 kl. 18:50 skrev Vermaas, Joshua <Joshua.Vermaas_at_nrel.gov<mailto:Joshua.Vermaas_at_nrel.gov>>:

Hi Per,

I think atom groups are useful to think about here, even though fftk doesn't believe they exist. Basically, each CH2 group in your fatty acid alone will be charge neutral. The ester or whatever other linkage will also be close to charge neutral. So whatever charge is on the deleted atoms (hopefully just a hydrogen?) would get added onto the attachment point. For systems I've worked with before, this tends to be a carbon, which will have a charge very close to 0 in the final result.

To get around the unparameterized bonds, angles and dihedrals, how bad would it be to cut the fatty acid down to an ethyl or propyl group? In that case, you would definitely assign fixed 0.09, -0.18 and -0.27 to the hydrogens and carbons, but you'd be able to explicitly scan the dihedrals you'd be missing.

-Josh

On 2019-09-19 06:16:35-06:00 owner-vmd-l_at_ks.uiuc.edu<mailto:owner-vmd-l_at_ks.uiuc.edu> wrote:

Hi all,

I am trying to use the ffTK to parameterize a rather large (41 atoms) molecule, which consists of a salicylamide part and a hydro-carbon chain ending with a carbonyl group (so essentially a fatty acid) bound to the nitrogen of the salicylamide.
Aside from being quite large, so the QM calculations take quite some time, I also have issues with placing the water molecules properly for determination of interactions energies, presumably due to steric clashes with nearby atoms.

I did submit the molecule to Paramchem, and the high penalties (43 at most for a dihedral, 16 for the charges) are all found in the salicylamide part of the molecule. Therefore I think it makes sense to split the ffTK parameterization, working on the salicylamide and essentially leaving the fatty acid tail as it is.
However, what then becomes the strategy for eventually joining these two fragments at the end? If I do get optimized charges for example, those will then sum to zero, right, but when I add in the fatty acid part, the total sum is not going to be zero again, most likely? Do I include this by saying that the hydrogen atom which is lost upon joining the fragments should bear the full charge sum of the fatty acid tail? And how about bonds, angles, dihedrals that cross between these two fragments?

Many thanks in advance
/Per