[gmx-users] Simulations with solvent shell

[Mark Abraham]

Aditi Borkar wrote:
> Dear All,
> 
> I created a simulation box in editconf with -d 2.5. Then in genbox, I
> used the -shell 1.4 option to define a 1.4 nm thick layer of solvent
> around my protein. When I visualized the system in Rasmol, as
> expected, there was a lot of "empty space" in the simulation box.
> Should this vacuum create any artifacts in PBC or other energy
> calculations? I did not receive any warnings or messages and so
> continued with the MD simulation.

Why did you define a spherical shell inside a periodic box? This will 
get the worst of all worlds - periodicity artefacts, finite-size effects 
and boundary effects.

> During the MD, I saw that some ( <10) water molecules travel away from
> the shell/layer around the protein and come to lie in the vacuum of
> the box. 

That's normal.

> Correspondingly the box size also varies depending upon the
> presence of such water molecules. After about 60 ps, the box reduced
> in size to just accommodate the solvent shell around the protein. And
> as yet the solvent forms a layer around the protein and is not
> distributed evenly in the box. By 100 ps the solvent molecules no more
> form a layer but are distributed evenly within the box (cube)

The box will only change size if you are using pressure-coupling, but 
you've not told us about this.

> One last observation is that when I viewed this trajectory in VMD, the
> molecule seemed to translate in space. There's no option to view the
> unitcell in VMD, so I do not know whether the whole box is shifting or
> just the protein+solvent is shifting inside the box.

The whole box can't shift... it's periodic, and there's no preferred 
unit cell. If you'd prefer a particular arrangement of box and atoms 
then you need to read trjconv -h, consider the various options and 
experiment until you're happy.

> Please let me know whether these observations are acceptably normal.

It seems very unlikely that you have both the simulation you hoped you'd 
have and a successful simulation :-)

Mark