[gmx-users] Smaller Area Per Lipid for DPPC Bilayer
t.piggot at soton.ac.uk
Wed Sep 12 21:51:42 CEST 2012
I'm not so sure as to say that all of the parameters for simulating
lipid membranes were not developed with a dispersion correction. The
Berger parameterisation used a dispersion correction for the pentadecane
simulations which were used to parameterise the tails (although as far
as I can tell from the Berger paper, this correction was not used in the
DPPC membrane simulations). Furthermore, the GROMOS 53A6 parameters of
Kukol were tested using simulations which applied a dispersion
correction (although you could argue that these GROMOS parameters were
initially developed without this correction) and if I remember correctly
the CHARMM27 (but not CHARMM36) lipid parameters were intended to be
used with a dispersion correction applied (although these parameters are
not for use with NPT simulations).
I would still argue that above all else, you should choose parameters
that someone has shown to accurately reproduce the experimental membrane
properties, irrespective of whether that is the original
parameterisation work or not (it may well just be your own simulation
tests). The Berger force field is a good example of where this sort of
testing/validation has been important. Several papers have shown that
PME should be used with this force field and not the direct 1.8 nm
coulombic cut-off used by Berger et al. Furthermore, in our work I
mentioned before, we show that with a 1.0 nm cut-off and no dispersion
correction (so the van der Waals parameters I believe were used in the
Berger DPPC simulations) there are several membrane properties that do
not match the experimental range. I do agree with you for the example
here though, it seems (from the information provided) the dispersion
correction should not be included with 1.2 nm cut-offs (and this also
agrees with results from three different cut-offs tested with the Berger
force field in our work).
On 12/09/12 19:30, Christopher Neale wrote:
> While dispersion correction is a great idea that helps to reduce the impact of the precise choice of cutoff distance on the results, the Berger parameters (and indeed all other parameters) were not developed with the inclusion of dispersion correction and one could argue that it is thus non-optimal to include dispersion correction here... especially since it leads to poorer results.
> This is separate from the difference between isotropic dispersion correction and a proper "PME"-type LJ term. Both of which are expected to lead to smaller APLs.
> When using anisotropic pressure coupling for lipid bilayers, you should use 1 atm in all dimensions.
> -- original message --
> Right I guess my biggest concern was diffusion. I did in fact do 12
> simulations of DPPC bilayers for 100 ns each, and still got the
> aforementioned APL and diffusion. When I turn off the dispersion, I
> get more appropriate APL and MSD values, that match other papers, even
> when only looking at one simulation. To me, it does not seem the mdp
> file I used is able to get more common APL and diffusion values even
> when averaging over a large number of simulations.
> As for the pressure in the x/y direction, is it more appropriate to
> use 1 atm or 0 atm for bilayer simulations?
Dr Thomas Piggot
University of Southampton, UK.
More information about the gromacs.org_gmx-users