[gmx-users] Re: Setting up a complex membrane simulation

[Christopher Neale]

The simple answer is yes, you could make Lipid_A-box.gro larger in the bilayer plane. That probably won`t
address the underlying problem though.

As far as I know, genbox doesn't take periodicity into account. That means that with larger species, such as lipid A
you are going to need to start with a much larger box and let pressure equilibration bring it down to the correct 
size in the context of PBC during mdrun. The alternative is to build a crystal with your own script and then set the
box boundaries yourself so that periodicity is taken into account.

Note that this is a major limitation of genbox and it would be great if somebody had the time to address it...

Also note that there will be a similar packing problem between lipids even if you discount problems with PBC
packing. There are only 2 ways to deal with this: (a) start with a crystal or (b) start with a sparse bilayer and have a
good equilibration technique. If you go for option A then you need to be aware of biases from the starting 
crystal. If you go for option B, then you'll likely have problems with lipid A acyl chains moving out toward bulk water 
and then not equilibrating on your achievable simulation timescale (there's at least one paper out there describing 
how to build a lipid A system and circumvent some setup problems, you should read it. I think it's at least 10 years 
old). The main problem with setup is that lipid A equilibration times are orders of magnitude larger than those for
phospholipids. I have personally had good success with high-temperature equilibration in which I add restraints to
a number of atoms that keep their z-coordinates in certain regions (where z is along the bilayer normal).

In fact, a good technique is probably to build a sparse lipid A bilayer,
restrain all Z coordinates of all lipid A atoms to their original values, and run some high-temperature MD to get 
some preliminary packing before slowly releasing the z-value restraints. I haven`t done this myself, but it is what I
would do the next time I need to build a lipid A bilayer from scratch.

Finally, there are papers recently in the literature of lipid A simulations. Perhaps you can get a lipid A bilayer from
one of those groups. I didn't have any luck obtaining these myself but perhaps they will be kinder to you if you ask 
nicely. Then you can bypass the building step entirely.

Chris.

-- original message --


Ok, what i've currently done:

I start with a pdb file of a single molecule Lipid_A and its topology and
put it in a box:

editconf -f Lipid_A.pdb -o Lipid_A-box.gro -bt triclinic -d 1.0

I want a total of 128 molecules of this lipid in my system so I use genbox
to add 127 more:

genbox -cp Lipid_A-box.gro -ci Lipid_A.pdb -p Lipid_A.top -o 128_Lipid_A.pdb
-nmol 127

This only ads 8 molecules, I'm going to guess because more don't fit in my
box. How do I deal with this? How can i know in advance how big i have to
make the box?

kind regards,