[gmx-users] A. - Vol. / lipid & associated errors

[Erik Lindahl]

Hi,
> I'm convinced that you guys, before a production run of a bilayer 
> system,
> perform an equilibration to get the feeling about the area/volume per 
> lipid.
>

Definitely. Unfortunately it's not quite so good that we equilibrate 
until a certain predetermined metric has been fulfilled, but rather: 
equilibrate as long as you can afford. Yes, it's stupid in one way, but 
the problem is that no matter how long you equilibrate you can always 
find some properties that require longer timescales to reach 
equilibrium... (think lipid flip-flop on second timescales).

> Can someone give some numbers about the associated property errors 
> between the
> MD obtained values and the experimental ones? 5 % is good in both 
> properties?
>

That depends on your forcefield. First, volume/lipid is much more 
well-defined experimentally and show lower fluctuations. However, most 
non-lipid forcefields are up to 20% off on both measures. We developed 
some lipid-specific force fields about 5 years ago that have less than 
1% error for common lipids, but of course that's not quite comparable 
since we're obviously parameterising for the lipids.

In general, I would recommend using a lipid-specific force field if you 
are interested in examining and reproducing properties of pure 
bilayers.  On the other hand, if your eventual goal is to use the 
membranes as a solvent for proteins you will probably be quite happy 
with setting the area/lipid to the correct value and the using whatever 
force field you like.

Cheers,

Erik