On Thu, May 5, 2011 at 9:03 PM, Sasha Buzko <obuzko.ucla.edu> wrote:
> Hi all,
> I'm running simulations of small proteins in mixed solvent. The very
> common problem I'm seeing is the protein being out of the solvent box
> after the post-simulation image operation. Since I'm looking for protein
> surface solvation patterns, this makes life somewhat difficult.
>
Centering and imaging *can* be a tricky business in some cases. Here is a
(as I've seen so far) sure way to image "properly". sander/pmemd do
pressure scaling/wrapping on a per-molecule basis, which means that only
whole molecules get imaged. Molecules are assigned by leap as any entity in
which all atoms in that molecule are covalently bonded to at least one other
atom in the molecule, and NOT covalently bonded to any atom that is in a
different molecule. Keeping this in mind, simply center each solute
"molecule" and image everything around it sequentially.
That is -- suppose we have 4 solute molecules surrounded by solvent, each
molecule is 25 residues large. The center/image ptraj script will look like
this (using familiar for truncated octahedrons and omitting it for
rectangular cells):
center :1-25 mass origin
image origin mass (familiar)
center :1-50 mass origin
image origin mass (familiar)
center :1-75 mass origin
image origin mass (familiar)
center :1-100 mass origin
image origin mass (familiar)
Note that the :1- is intentional in each case. The first one will center
the first solvent molecule and image everything around it. What this
assumes is that it images the second molecule (residues 26-50) appropriately
next to the first molecule if it had been wrapped during the course of the
simulation (or leaves it alone if it didn't change boxes). Now that the
first 2 molecules are in the same box, we center the first 2 molecules and
image everything around them, then use the same reasoning for the remaining
molecules we want at the center of the box.
Note that in most cases only the first 2 center and image commands will give
you the desired behavior, but better safe than sorry :).
Is there a way to stack several of these unit cells to get a complete
> solvation picture for the protein?
Yes. In VMD, open up the Graphical Representations tk window, then click on
the "periodic" tab. The rest should be self-explanatory.
> Alternatively, is there a way to keep
> the protein within the bounds of the box.
>
Yes, 2 ways, but there are problems with both. Run with iwrap=0. Then,
imaging is as easy as
center :1-100 mass origin
image origin mass familiar
I can provide a more detailed explanation of why this is the case if it's
not that clear (but I'm lazy now :)). However, iwrap=0 has problems (at
least as long as we have ASCII formatted restarts) for long simulations.
The other way is hacking the prmtop file. If you figure out how to combine
all solute molecules into a single molecule definition (see the sections
ATOMS_PER_MOLECULE and SOLVENT_POINTERS and their descriptions on
http://ambermd.org/formats.html for details), then pmemd and sander will
*never* image them separately. However -- if you do this wrong, you will
pay for it (with segfaults and other difficult-to-trace errors -- I just
recently spent ~1 hr determining that a segfault thrown by pmemd was caused
because the ATOMS_PER_MOLECULE section of the sample prmtop was wrong,
though I was not the one to modify it).
And this concludes more than you probably ever wanted to know about imaging,
wrapping, and molecules in the amber topology.
HTH,
Jason
--
Jason M. Swails
Quantum Theory Project,
University of Florida
Ph.D. Candidate
352-392-4032
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Thu May 05 2011 - 14:30:04 PDT
