[gmx-users] About simulating lipid bi-continuous cubic phase

[Dallas Warren]

George,

> I am attempting to simulate lipid cubic phase using Coarse grained MD
> from the spontaneous aggregation approach, starting from random
mixture
> of lipids in solution, in the spirit of JACS 2009 paper from the
> Martini
> community. I had several technical issues before setting up the
> simulations and I would appreciate expert opinion on these.

I can't comment in terms of the coarse grain type of MD, as we are
performing united atom simulations of phase behaviour, but the
principles are still the same.

> My main concern is about the interplay between the number of lipids in
> the system, initial size of the simulation cell and the pressure
> coupling method to use. What should be the recipe for choosing the
> initial number of lipids? Obviously, I would think that one has to
> choose the number large enough to form at least one unit cell of the
> bi-continuous cubic phase. But what happens if I don't choose exactly
> this "magic" number, but slightly more? Then I would think that after
> running the simulation, the simulation box would contain more than
just
> a unit cell of the cubic phase. Can this be a problem?

It is good that you are concerned, as it does influence the result.  If
you have sufficient size / number of molecules to form a non-integer
number of unit cells, then the final structure will be distorted from
what you would expect as the end result.  The only way to get beyond
this is have the system large enough that it is not an issue, and
computational speeds are not at this level as of yet.  You have a number
of options available.  Put in the "correct" number or dimensions, but
you are constraining the results of the simulation, you might as well
preassemble it, since you aren't actually allowing it to spontaneously
form a structure, you are guiding it.  Acknowledge that the structure
formed may not be perfect and will have flaws in it, due to the
dimensional issue, and simply work that into your work.

> Also does the initial size of the simulation box matter if one
> simulates
> spontaneous aggregation process as long as the molecules initially
> "fit"
> into the box?

If you are going for spontaneous aggregation, then random distribution
of the molecules within the box is what you should be starting from.

> And finally, what role should isotropic vs anisotropic pressure
> coupling
> play in such approach? Which pressure coupling method should be
> preferred in such spontaneous aggregation simulation?

We played with this for quite some time.  In a perfect world, it would
be best to using anisotropic pressure coupling, so you do not constrain
the structure that is formed.  However, in the world of MDS, it is
simply not possible.  You encounter major issues with simulation
stability, the way the pressure coupling and alternating charges across
one dimension of the box interact means that eventually you will end up
with the box distorting to the point that you end up with a match stick
or a pancake.  We even played with adding various tweaks to the coupling
algorithms that increase the resistance to changes in the box dimensions
as they deviate from cube or get too close to the point at which it will
explode.  The algorithms works great.  But you get back to the same
point again, ultimately you are constraining the end result.  And the
flow on effects of that are harder to quantify than simply staying with
isotropic.

We are putting together a publication on the phase behaviour stuff we
have done, some mention will be made of these issues in that.

Catch ya,

Dr. Dallas Warren
Medicinal Chemistry and Drug Action
Monash Institute of Pharmaceutical Sciences, Monash University
381 Royal Parade, Parkville VIC 3010
dallas.warren at monash.edu
+61 3 9903 9304
---------------------------------
When the only tool you own is a hammer, every problem begins to resemble
a nail.