[gmx-users] Re: Your Question about unwanted waters
dmobley at gmail.com
Mon May 8 18:53:12 CEST 2006
For the record (since I'm sending this to the users list also), the
problem was that I had a protein with some small "pockets" in it that
are empty in the crystal structure. Some of these pockets were just
large enough that a water molecule could fit in when genbox solvates
the structure, even though no water molecule should be there. In
subsequent simulations, these waters would often remain trapped inside
the protein and drift around, exhibiting very long correlation times.
I finally "solved" the problem by modifying the vdW radii used by
genbox for solvating the protein. The way this works is that genbox
uses the vdwradii.dat file to read atomic vdW radii for solvation. It
will search first in the current directory, then in the GROMACS dirs.
So I created my own vdwradii.dat and used 3.0 A for all of the radii,
forcing GROMACS to leave 3A between the protein and any water
molecules, which meant no waters were put inside my protein. Of
course, it also meant that water molecules were missing from a 3A
shell around the protein. So then I ran a long molecular dynamics
"equilibration" step (about 1 ns, which was probably overkill) with
the protein restrained to allow the water around the protein to
equilibrate, before even running "regular" equilibration. It's
probably worth pointing out that my production calculations are up to
5 ns, and I never see water enter those sites that genbox was putting
water into on 5 ns timescales, which is reassuring or troubling
depending on how you look at it.
Anyway, that's what I did. Not particularly elegant, but it works.
However, I would emphasize that you should be *sure* where you want
and don't want water if you are going to use this approach.
On 5/8/06, Arneh Babakhani <ababakha at mccammon.ucsd.edu> wrote:
> I was wondering if you ever figured out the problem that you posted here:
> I'm having the same issue (solvating a membrane, and it's placing waters
> in the core of the lipid bilayer).
More information about the gromacs.org_gmx-users