[gmx-users] Re: Gromacs 54a7 force field

Thomas Piggot t.piggot at soton.ac.uk
Thu Jul 19 01:36:44 CEST 2012


Yes, I agree with you regarding the combination of Berger and GROMOS 
force field and requiring validation. I just wanted to point out the 
interactions between the protein and lipid are treated in the same way, 
irrespective of the different GROMOS protein force field used (when 
using the parameters from lipid.itp).

Further on the topic of validation, the only work I know of that has 
studied in detail the Berger/GROMOS combination 
(doi:10.1088/0953-8984/18/28/S07) says that when using the standard 
method as employed by the parameters in lipid.itp:

"the strength of interactions between hydrocarbon lipid tails and 
proteins is significantly overestimated, causing a decrease in the area 
per lipid and an increase in lipid ordering"

So the standard approach is probably not the most appropriate way to 
combine the Berger lipids and the GROMOS protein force field. Rather 
using combination rules (as for example done in the Berger/OPLS 
combination) is probably better.



On 19/07/12 00:15, Justin Lemkul wrote:
> On 7/18/12 6:57 PM, Thomas Piggot wrote:
>> Hi,
>> Justin, I am interested by your comments regarding the CHARMM lipids. In
>> particular can you elaborate as to why you think that the CHARMM 
>> lipids are
>> better than the united-atom ones (such as Berger and several GROMOS 
>> variants).
> I think there's nothing wrong with Berger lipids, and I myself use 
> them routinely.  They do a decent job of reproducing lipid behavior, 
> no question.  My own use is motivated by historical reasons primarily 
> since I began work with them many years ago (also around the time that 
> 53A6 was all shiny and new) and I'm trying to keep consistent.  It 
> seems to me that the CHARMM36 parameters are very thoroughly validated 
> and are very modern.  The Berger lipids were developed a long time ago 
> using very short (by modern comparisons) simulations and old parameter 
> sets that have since been improved upon.  There are numerous 
> reasonable lipid models out there, to be sure.
>> As for the original question, the modifications in going from GROMOS 
>> 53A6 to
>> 54A7 will not influence the combination with the Berger lipid 
>> parameters, if the
>> most common approach of using the parameters from the 'lipid.itp' 
>> file is taken.
>> The interactions between protein and lipid will remain the same, with 
>> the van
>> der Waals interactions between the protein and lipid treated using 
>> the GROMOS87
>> parameters as defined in lipid.itp. From my experiences I would strongly
>> recommend using 54A7 over 53A6, as we have seen instability in short 
>> helices in
>> 53A6 that is not reproduced when simulating the same systems with 
>> several other
>> force fields.
> I agree with this, I'm just a stickler for validation.  A test system 
> using this force field combination should certainly replicate some 
> known behavior.  It's generally accepted that any Gromos derivative is 
> compatible with the Berger parameters, but I've never seen anyone 
> demonstrate it systematically aside from their own usage for 
> particular cases.
>> For the 'best' force field to choose when simulating a 
>> membrane-protein system,
>> there is no definitive answer that I (or anyone else) can give you 
>> (yet). My own
>> opinion is that currently CHARMM36 is probably too slow (given that I 
>> would
>> strongly recommend the use of the CHARMM TIP3P water model with this 
>> force
>> field) and that the all-atom protein force fields are probably better 
>> than the
>> united-atom ones. This means that I would (for a PC membrane) use 
>> Berger for the
>> lipids with OPLS-AA/L or an AMBER force field for the protein. This 
>> is just my
>> (current) opinion though, I strongly suggest doing lots of your own 
>> reading
>> before making your mind up.
> Indeed.
> -Justin

Dr Thomas Piggot
University of Southampton, UK.

More information about the gromacs.org_gmx-users mailing list