[gmx-users] Force Field for Bilayer Simulations with Cholesterol and Proteins

[Justin Lemkul]

On 5/5/14, 12:39 AM, David Ackerman wrote:
> Hello,
>
> I have been performing simulations of multi-component bilayers using
> parameters described in the following paper:
> http://www.ncbi.nlm.nih.gov/pubmed/23470157. They use the ffgmx forcefield
> with Berger lipid parameters included. They also use cholesterol from
> Holtje et. al (
> http://www.sciencedirect.com/science/article/pii/S000527360100270X), which
> was parameterized for the ffgmx force field.
>
> I noticed that unlike the KALP-15 GROMACS tutorial which suggests deleting
> the default lipid-gromos Berger interactions, the authors kept those
> interactions in the force field. In light of newer force fields, are these
> older force fields wither Berger parameters still acceptable? The
> cholesterol model seems commonly used, so I am unsure how to use
> cholesterol with a newer force field.

Just to be clear, the my tutorial instructs the user to remove those 
interactions because they are force-field specific and otherwise override the 
nonbonded interactions produced by the normal combination rules.  If one seeks 
to use a Gromos96 representation of the protein + Berger lipids, leaving those 
interactions in place is totally invalid.

>
> Furthermore, if I were to include proteins in the simulation, would I still
> be able to use the same ffgmx force field with Berger parameters (including
> lipid-gromos interactions)?
>

I can't speak to the Holtje parameters, but I have taken other published 
Gromos87 cholesterol models and transferred them to Gromos96 (just a few atom 
types changed); the results of those simulations (original Gromos87 vs. 
Gromos96) are essentially identical.  It is a very simple validation to do.

My general advice is that people shouldn't be using ancient force fields with 
undocumented ad hoc changes, as is the case with gmx.ff; that is a major reason 
why it has been removed from Gromacs in the 5.0 series.  Newer, better, force 
fields exist, and all indications are that these Gromos96 parameter sets can be 
combined with Berger lipids with trivial modifications.  Of course, the larger 
debate exists over whether such a combination is the best model, but that's a 
different discussion :)

-Justin

-- 
==================================================

Justin A. Lemkul, Ph.D.
Ruth L. Kirschstein NRSA Postdoctoral Fellow

Department of Pharmaceutical Sciences
School of Pharmacy
Health Sciences Facility II, Room 601
University of Maryland, Baltimore
20 Penn St.
Baltimore, MD 21201

jalemkul at outerbanks.umaryland.edu | (410) 706-7441
http://mackerell.umaryland.edu/~jalemkul

==================================================