Date: Tue, 6 Mar 2001 23:01:34 +0100 (MET)
From: Thomas Huber
Subject: Re: lipid bilayer MD simulations

Dear Colleagues,

I have tried to develop an all-atom force field for POPC following the
Cornell et.al. procedure with a 6-31G*/RESP parametrization of
multimolecule and multiconformer POPC fragements.
The crucial point in this work was the choice of the van der Waals
parameters. When I took the amber94 parameters, constant pressure
simulations resulted in very small area per lipid values and high
densities indicating a gel phase like inducing effect even at high
temperature.
After I changed the LJ parameters to the new CHARMM all-atom lipid force
field, the simulations behaved well.
I would strongly recommend to use the CHARMM forcefield parameters for
lipids (you can switch some compile time options in the amber code to
allow for the full CHARMM type forcefield, I can speak only for AMBER 4.1,
I haven't got the later versions).
It might be possible to combine the CHARMM/lipid and AMBER/protein
forcefield, they are not too dissimilar. But there colud be always
problems mixing different parametrizations.
If you want to do united-atom simulations, the GROMOS/GROMACS forcefield
parameters seem to be some sort of common sense, at least there are
several lipid/protein simulations with this forcefield. There are also
AMBER based united atom simulations, other than the ones you mentioned.
But I doubt, if they are all really realistic, in terms of allowing
accurate predictions of the area per lipid in constant pressure or
constant surface tension simulations.
The question of taking lipid parameters seems to me very similar to the
choise of a good water model. It depends on the questions you have to the
system, what accuracy you would need for a forcefield.
As my background is lipid bilayer membrane 2H NMR, my requirements are
very high, and now after about 5 years of work in this field, I have some
very good data to publish, but the last years....

Yours,
Thomas


On Wed, 7 Mar 2001, Xavier Deupi wrote:

> Hi all,
>
> I'm interested in performing MD simulations of membrane proteins in an
> explicit water/lipid bilayer environment using AMBER. I guess some of you
> are working in this field, so I have some questions for you...
>
> In AMBER distributions and home page, I have only seen mention to lipid
> bilayers in one of the examples for roar, where DLPE
> (dilauroylphosphatidylethanolamine) is used, with the associated DLP
> residue type. However, I haven't found any reference to this residue type
> in the parameter files. I guess people also use other models for the lipid
> bilayer (DPPC, POPC,...), so my question is obvious... are parameters for
> these phospholipids available anywhere?
>
> I have read that united-atom force-fields are quite common for such
> simulations. Is this correct? Anybody is using all-atom force-fields? If
> so, are parameters available?
>
> Finally, I would like to keep in touch and have some feedback with people
> working in this field... you know, it's very hard to learn on your own
> (although sometimes it's the best way) :-)
>
> Well, as always, thank you all in advance.
>