[gmx-users] simulation confined to a small region

[Mark Abraham]

> Has anybody had success on running a simulation on only a part of a
> protein? I have a very large protein, and I'm only concerned with
> analyzing the motions of two specific close sidechains rather than
> analyzing the complete system. Do you recommend that I set the simulation
> box to very very small dimensions and then orient the box so that the two
> side chains are in the center of the box? If that's how it is done, I've
> had no luck with using editconf. Any suggestions?

This problem is closely confounded by how you want to treat the region of
the protein you're planning to chop off. Periodic boundary conditions are
useless to you, unless you take enough of the protein that you think it's
electrostatic interactions with these sidechains is well-enough covered,
use position restraints as appropriate on that protein fragment, cap the
inevitable bits of sequence that will be dangling, and then take at least
two complete solvation layers around the whole lot - since the water
structure near the former "inside" parts of the protein will be weird.

An alternative is to do "shell MD", where you again pick a reasonable
chunk of your protein, cap bits of it, use position restraints and/or
fixed atoms, introduce a solvent layer around the sidechains, and then
another solvent layer around that with PR and fixed atoms, wide enough
that the electrostatic effect of the vacuum outside is probably
negligible.

In the former case, editconf will be no use to you until you've built the
system by hand with some other tool. In the latter case, I can't see
editconf being of use at all.

Mark