[gmx-users] lipid gel phase simulation model building

[Carsten Selle]

>
> Dear gromacs users,
> >
> > the gel phase of some lipids as the L beta prime of DPPC is
> > characterized by all-trans alkyl chains which are tilted with respect to
> > the bilayer normal. I would like to set set up a system analog to the
> > one described by Venable, Brooks & Pastor, J. Chem. Phys., 112 (2000),
> > pp. 4822. They used a periodic box which is cubic (39 x 39 x 63 A***3).
> > I tried editconf and genbox to  generate a cubic simulation box and a
> > suitable aqueous environment for the bilayer I generated from crystal
> > data. However, since the box is cubic, there is plenty of room for
> > unwanted water-alkyl interaction. genbox accordingly placed water in
> > hydrophobic places.  The water molecules should (almost) only interact
> > with with the lipid headgroups and water. Any helpful comment is greatly
> > appreciated.
> >
> > Best regards,
> > Carsten
>   
>
> Try keepbyz.sh at http://wiki.gromacs.org/index.php/Membrane_Simulations
>
> Also note that, assuming your membrane normal is along z, editconf  
> will make the box bigger than you would like in the xy plane since it  
> probably doesn't fit perfectly by PBC and water will come in (at the  
> sides especially) during initial equilibration. If your gel really  
> does do PBC already then try just cutting down the x and y and  
> checking it out via g_mindist to ensure the crash is not too bad in a  
> trial-and-error process. You could also try inflategro from the  
> Tieleman lab without using a protein and that may help you get the  
> sides, although I am not sure that this would work as expected.
>
> Chris.



Dear Justin and Chris,

Thank you very much for your kind advice. I am happy to follow your 
suggestions. This might be an exotic (merely nonphysiological) topic. 
Thus, I try to elucidate it in the following in order to make it more 
accessible for future discussions/search.

A section of the the L beta prime (gel) phase looks schematically like 
"drawn" in the following, with "o" being the polar headgroups and and 
"/" the hydrophobic tails.

 oooooooo
////////
////////
oooooooo

The attempt to establish a (rectangular) simulation box and a solvent 
environment resulted in the following, with "|" and "-" being the z and 
x or y box boundaries, respectively, and "*" representing the solvent 
molecules:

|----------|
|**********|
|**oooooooo|
|*////////*|
|*////////*|
|oooooooo**|
|**********|
|----------|

One observes that water really comes in at the sides, as Chris pointed 
out. Apart from that, the model is close to figure 1 (left panel) in the 
reference from Venable et al.(2000). As the chain contacts are very 
close in gel phases in general and specifically in the crystal-derived 
structure that I used, water is not introduced in interstices between 
the chains. The wiki article mentioned above seems to focus on the 
latter problem which is not that applicable in the case summarized here.

Best wishes, and thanks again
Carsten