[gmx-users] Lipid parameters for GROMOS96 force fields

[Justin A. Lemkul]

Krzysztof Mlynarczyk wrote:
> 
> 
> 2010/1/21 Justin A. Lemkul <jalemkul at vt.edu <mailto:jalemkul at vt.edu>>
> 
> 
> 
>     Krzysztof Mlynarczyk wrote:
> 
>         2. If not, is there any way to derive the proper parameters for
>         the force field of my choice using the lipid parameters from
>         Peter Tieleman's website or e.g. the parameters published by
>         Andreas Kukol for G53a6?
> 
> 
>     I don't see why you need to do such reverse engineering.  The Kukol
>     parameters for lipids under 53a6 can be directly combined with a
>     G53a6 protein without any issues; I believe that was the purpose of
>     the whole new derivation :)
> 
> 
> I received a message that G53a6 is beta-sheet biased and alpha helices 
> do not perform as well as they should. My protein contains 7 
> transmembrane helices, that's why I'm worried.

Is this published somewhere?  That would be important information.  Perhaps this 
is the case for model peptides or short fragments, but I have certainly done a 
number of simulations using 53a6 with well-folded globular proteins and I do not 
see any such instability (i.e., alpha->beta conversion or unwinding of 
alpha-helices).  I do believe it is possible in certain scenarios, but I don't 
know that a large 7TM protein like yours would suffer adversely.

> I know that there are changes between parameter sets both in non-bonded 
> and bonded terms and one rtp entry will probably not work well when 
> pasted into a different force field from the same family. G96 family 
> uses symbols like gd_5 that are substituted by appropriate parameters 
> later through the use of preprocessor. While it is possible to find that 
> gd_5 is the same as gd_15 in another version of G96 and substitute those 
> symbols in topologies, the changes in non bonded parameters still can 
> spoil what was working well elsewhere. That's why I was also asking for 
> some checked and ready-to-use topologies for a particular force field.
> 

Many of the bonded parameters carry over between force fields, but certainly new 
entries were created between 43a2 and 53a6, so yes, some re-working would likely 
be necessary.  There is a lipid 43a2 parameter set on the User Contribution 
site, like I said before, I just don't know if there is a reference for it.

> 
>     As an aside, you are quite right that multiple force fields within
>     the same simulation is incorrect.  However, the Berger lipid
>     parameters may be an exception to this rule, since they are really a
>     hybridized version of OPLS-UA and Gromos87 parameters (some of which
>     were modified anyway), so they really don't belong to any one
>     particular force field.  The Berger/G87 combination is widely used,
>     but essentially amounts to the following: lipid interactions are
>     Berger-Berger or OPLS-OPLS interactions, while protein-lipid
>     interations are Berger-G87, and protein-protein interactions are
>     G87-G87.  You can see quite quickly why things become complicated!
> 
>     Based on a discussion I had with Dr. Tieleman, it seems to be
>     reasonable to use the G96 parameter set of your choice in
>     conjunction with lipid.itp (Berger lipids), although other
>     approaches may be more rigorously correct (pure G96 parameters such
>     as those by Kukol, pure OPLS recently derived by Ulmschneider, or
>     the modifications to the Berger parameters from the Tieleman group,
>     to name a few).  If you want to use a G96-lipid.itp combination, I
>     created a tutorial that teaches you how to build the system and
>     properly prepare the topology.  It is linked from the Tutorials page
>     of the Gromacs site.
> 
> I found this tutorial earlier and was also in doubt if this approach was 
> correct. But if it works, perhaps I should give it a try.
> I gotta make a _good_ decision in the end...
> 

As do we all :)  My work with G53a6+Berger has thus far been quite reliable, 
from everything I can measure, but that certainly does not preclude the 
possibility (even likelihood) that there are better procedures out there, like 
those I quoted above, and certainly others (CHARMM is also popular for membrane 
proteins, but Gromacs will only *officially* support CHARMM as of version 4.1).

-Justin

> Christopher
> 

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

Justin A. Lemkul
Ph.D. Candidate
ICTAS Doctoral Scholar
MILES-IGERT Trainee
Department of Biochemistry
Virginia Tech
Blacksburg, VA
jalemkul[at]vt.edu | (540) 231-9080
http://www.bevanlab.biochem.vt.edu/Pages/Personal/justin

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