[gmx-users] about water molecules in popc

[Erik Lindahl]

Jose D Faraldo-Gomez wrote:

>
> If the problem is waters in the lipid tail region, you could either 
> remove them with a script, or have a local copy of genbox's input file 
> vdwradii.dat (or something like that) with increased values for the 
> radii of carbons in the lipid tails, so that the 2 leaflets 
> effectively overlap and no waters can be added there.

I did this a couple of years ago; there are essentially two different 
problems:

1. You don't want grompp to add any water in the space between lipids 
and protein.
2. During the run, you don't want water to diffuse into that space 
before the lipids have relaxed around the protein.

The first part is easy to solve by using a vdwradii.dat where you 
increase the radius of the lipid chain atoms.

To solve (2) I simply wrote a small perl script that removed all water 
molecules that had z-coordinates in the membrane. I also had some water 
that should be in a cavity in the protein, so I kept all water molecules 
that were closer than 1 nm (or something) to the XY center of the protein.

Sorry, can't find the script, but the algorithm was essentially:

If (z_OW < upper_headgroup && z_OW > lower_headgroup) {
# we are in the membrane
dx = x_OW - x_protein_center;
dy = y_OW - y_protein_center;
if((dx*dx+dy*dy) > max_distance_from_protein_center) {
# remove it
}
}

Then i ran the system for a short time (say 10 ps), removed water & 
updated the topology, and repeated this until no more water was entering 
the space.

If you use a smarter algorithm for placing the protein in the membrane 
in the first place you will have less trouble :-)


>
> If the problem is the interface with the protein, Graham Smith and 
> myself developed a methodology to optimise the lipid-protein contacts 
> in membrane protein simulations. The reference is
>
> European Biophysical Journal
> Eur Biophys J (2002)31:217?227
>
> E-mail me (or Graham) if you want the source code (it works for 
> Gromacs 2 only).
>
It's probably close to trivial to adapt for 3.0 in case you need it... 
There are only a couple of minor modifications to the library calls.


Cheers,

Erik