Potential energy and good parametrization - few questions

GROMACS version: 2018.4
GROMACS modification: No
I have a problem. I created DGDG membrane and used gmx energy to check potential energy (after energy minimization and after equilibration). The problem is that I use old parameters and new parameters, which I found in my old colleague’s folder. The old parameters give my potential energy -2,510^-5
The new parameters give me
-2,43
10^-5

But I think new parameters are more appropriate?
ladunkiDGDG_stare

I underlined differences in old parameters. For me C1 should have a higher charge (0.300 instead of 0.170) because it is between two electronegative Oxygens and C in glycerol instead of 0.280 should have a lower charge.

  1. Which parameter I should use old or new?
  2. Lower potential energy = better parametrization??? I thought that I probably need to check some membrane parameters like area per lipid/width of membrane/molecular order parameter etc. to know that.

I found also some publications about that, this one describes some charges
https://onlinelibrary.wiley.com/doi/epdf/10.1002/(SICI)1096-987X(199712)18%3A16<1955%3A%3AAID-JCC1>3.0.CO%3B2-L

Thanks in advance for all your advice.

On what basis did you make the change? There are typically rules in force fields about transferability and restrictions on H charges for compatibility. Ad hoc changes are generally not appropriate.

Not at all, and actually the total potential energy is entirely irrelevant. Force fields target specific conformational potential energies (relative, not absolute) and compare them against QM data, then condensed-phase properties. Charges are typically calibrated against molecular dipole moments, electrostatic surface potential, water interactions, etc.

If I were a reviewer of your work, I would be highly suspicious of what you were doing. Force field parametrization must be approached extremely carefully. Garbage in, garbage out. Unless you have some demonstration that your new parameters are more suitable for reproducing carefully chosen target data and molecular properties, don’t use them.

2 Likes

Thank you so much for the information. Yeah, I think I should be more careful and find a source of both of these parameters or find better parameters.

I know that for galactose residue, he get parameters from here OPLS all‐atom force field for carbohydrates - Damm - 1997 - Journal of Computational Chemistry - Wiley Online Library
(Figure 5)
dgdg
In this figure, I see that charge for C1 should be 0.300

but I don’t know from where he found parameters for glycerol.

And how to combine them? If you have parameters from one publication for glycerol and from another publication for carbohydrates and if you put these charges into your topology and you will have a non-zero charge on your molecule (DGDG should have total charge = 0), what then, what to do to have total charge = 0 if you have for example 0.270 or -0.170 total charge etc? Try to make parametrization for the whole molecule?

I also hear about topology generator, but I heard that this is not the best option like LigParGen Server (yale.edu)