Re: Setting up a membrane simulation

From: Bjoern Olausson (
Date: Fri Sep 16 2011 - 07:13:56 CDT

Hi Eric,

Minimization should always be the first step - to get rid of bad contacts (You
might want to minimize as long as you see a significant change in the
potential energy)

You want to use constraints to keep the water molecules away from the
hydrophobic core (but do not FIX waters), keep the lipid tails in place and
the headgroups to maintain the proper area/lipid ratio.
Have a look at the protocol CHARMM-GUI ( just build a
little membrane and look into the input files) uses.
When you set up these constraints you can heat your system and slowly decrease
the constraining forces. Once you are trough with this you are free to choose
whatever ensemble you like for the production run. You should discard the
first hundreds of ns without constraints from the production run.

NPgT only makes sense in my eyes if you have experimental results like order
parameters to validate you simulation against (run multiple simulations with
different SurfaceTensionTarget settings until one yields the same
orderparamter as from your experiments).
If you have to use NPgT depends on the forcefield you are using. Some should
yield the proper Area/Lipid ratio (e.g. CHARMM C36) some do not (e.g. CHARMM

If you know the lipid/area ratio you can fix the area by running a NPaT
simulation (UseConstantArea or useConstantRatio). make sure your membrane is
aligned to the "membrane normal Y)

Regarding PME:
I would choose PMEGridSpacing 1 for the entire equilibration runs since the
box size might fluctuate. Once you have a stable system you might want to nail
down the PMEGrid. You can fire up a short simulation with PMEGridSpacing 1 and
NAMD will calculate effective values for you to choose for PMEGrid{X;Y;Z} ;-)

Regarding the NVT ensemble:
If you want to calculate the membrane pressure, you should run one. Otherwise
the NPT ensemble is more Lab-like. The main difference is, that in NPT the
unit cell size is allowed to change, whereas in the NVT the size (Volume) is


On Friday 16 September 2011 09:52:05 you wrote:
> Hi all,
> I use NAMD 2.8 to simulate a pure DMPC membrane (72 lipids and no protein
> embedded) surrounded by water molecules. Up to now the trials I did failed
> with errors such as:
> ERROR: Constraint failure in RATTLE algorithm for atom 13561
> Looking around on the lists I found some hints but I also got frightened
> with all the possibilities/parameters that can be tuned (e.g.
> PMEGridSizeX,Y,Z vs PMEGridSpacing, NPT/NVT vs NPGT, Berendsen vs
> Nose-Hoover, use or not SurfaceTensionTarget, useConstantRatio,
> useConstantArea ...), wondering finally if a single bad choice could mess
> up my simulation.
> I tried two simulation protocols:
> 1) Minimization: 10000 steps
> 2) 100-ps NVT simulation with lipid head groups and water fixed in
> order that the lipid tail can reorganize 3) 100-ps NVT simulation with
> whole lipid fixed in order that water can reorganize 4) Equilibration: 1
> ns NPT (Langevin piston barostat)
> 4) Production: 100 ns NPT (Langevin piston barostat)
> and
> 1) Minimization: 10000 steps
> 2) Heating: NPT (Langevin piston barostat) from 0K to 300K by
> 100ps-steps of 10K 3) Equilibration: 10 ns NPT (Langevin piston barostat)
> 4) Production: 100 ns NPT (Langevin piston barostat)
> which one looks the more reasonable (if there is one ...)?
> I would also have a few additional questions:
> - I read in the NAMD membrane tutorial that in the early steps of the
> protocol, one should introduce a step where only the tail of the lipid can
> move Is that something very important to consider, especially in my case
> where there is no protein ?
> - some people use NPGT via the SurfaceTensionTarget option. Is a
> standard NPT simulation really so bad for this kind of simulations ? If I
> have to consider this option, how to set the value ? Is there some kind of
> standard one (e.g. 60) ?
> - my first protocol starts NVT pre-equilibration with an equilibration
> in NPT. Now, tt seems to me that NPT simulation would be better in the
> sense that my system would always breathe adapting its volume to its need.
> What do you think ?
> - should I prefer the PMEGridX,Y,Z options to PMEGridSpacing (e.g 1.0)
> ?
> Sorry in advance for the long and perhaps naive email but with 3-weeks per
> simulations, testing by myself all these parameters would take quite a
> while !!!
> thank you very much
> Eric

Bjoern Olausson
Martin-Luther-Universitšt Halle-Wittenberg 
Fachbereich Biochemie/Biotechnologie
Kurt-Mothes-Str. 3
06120 Halle/Saale
Phone: +49-345-55-24942

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:57:45 CST