[gmx-users] pressure coupling for lipid bilayer

[Erik Lindahl]

Hi,

>
>Yes, the lipid forcefield parameters should reproduce all the properties
>of the membrane. But because, in the case of the surface area, the limitation
>
>in size of the patch you can simulate introduces a systematic scrinckage
>of your lateral box dimension (ref: SE Feller and RW Pastor, Biophysical J.,
>1996 vol 71 p 1350), you have to apply a surface tension that rectify this.
>The forcefield is not the problem, the size of the system is the problem.
>
>
Both Scott (Feller) and Rick (Pastor) are nice guys and very good 
scientist, but I think
most membrane people now agree that the argument in the above article is 
NOT valid.

There are small finite size effects on bilayers (We've simulated from 64 
to 1024 lipids, see
Biophysical Journal, 79, 426 (2000)), but we're talking about 1-2 nm^2 
per lipid here, never
10 nm^2 per lipid.

The main problem isn't the box size but the forcefields used. If you 
look at the CHARMM
parameters used by Scott/Feller and e.g. the Gromos96 ones, you will 
find that the density
of a system consisting of only long hydrocarbons like pentadecane (15 
units) is 10-20% too low.

Of course, if you use these parameters for the chains in a membrane the 
volume/lipid will be too low,
and since the net surface tension (i.e. equilibrium area/lipid) is a 
very delicate balance between
headgroup and chain region (JCP 113, 3882 (2000)) it is no surprise that 
the area/lipid comes
out too low too.

Although you can argue that it is difficult to measure the area/lipid, 
the volume/lipid is easy
to measure to within 1% for most lipids, but even this value is way off 
with the forcefields above.

Understand me right - I don't think you make any terrible errors by 
simulating at a fixed area/lipid or
by applying a surface tension, but the main reason to do so is a 
deficiency of the force field
parameters, not scaling effects. If you want to solve it properly you 
can reparametrize the atoms
in the carbon chains to make sure the density is right for longer 
hydrocarbons; if you check the
contributed forcefield page at www.gromacs.org you can find such a set 
of modified parameters for
the Gromos96 forcefield. There is also a newer version of Gromos96 
parameters from the van Gunsteren
group that should be good, but I haven't tried those myself.

Cheers,

Erik