On Tue, Jan 22, 2013 at 10:55 AM, Kyle Roberts <ker17.duke.edu> wrote:
> Hi,
>
> I am attempting to run a steered MD simulation (using sander with
> ncsu_smd) where I pull on a loop of the protein in a given direction.
> I figure using a DISTANCE variable is the easiest way to do this, but
> unfortunately there are no two atoms that are lined up in the
> direction that I would like to pull. Therefore, I would like to add a
> dummy atom so that I can pull the protein loop in the direction of the
> dummy atom.
>
> Based on a previous thread on the list, I added a dummy atom by adding
> an ion to my protein and then editing the ion parameters (with parmed)
> to set the VDW, charge and GB Radii and Screen terms to 0.0. However,
> when the GB Radii and Screen terms are zero the overall energy is
> always "NaN". If I only change the VDW and charge parameters, my dummy
> atom still interacts with the environment (I am using implicit
> solvent) and slowly moves away from the protein during the simulation.
>
> My current solution is to restrain the movement of the dummy atom so
> that it cannot move away from the protein, however this is non-ideal.
> I assume that another solution would be to use explicit solvent,
> although I would prefer to stick with implicit solvent. Is there
> another way to use a dummy atom with implicit solvent? Alternatively,
> is there a better way to pull the protein in a specific direction
> without the addition of a dummy atom?
>
As long as the charge and the VDW parameters are zero and no bonds exist,
the ion is _not_ interacting with the rest of the protein. All GB energy
terms in the pairwise summation include a product of the charges (which is
0 with the dummy atom). Furthermore, as long as the dummy atom is at least
rgbmax away from the surface of the protein, it won't be included in the
calculation of the effective radii for any protein atoms, either.
I think what you are observing is either due to the thermostat providing
the dummy atom with velocity (especially if you are using a stochastic
thermostat) or natural translation of the protein. This translation is
periodically removed, and since the ion does not affect the center of mass
of the overall system very much, it will just appear like the protein is
fixed in space while the ion moves around it, when it may actually be the
opposite. Look at the nscm variable and how you can set it to NOT remove
COM motion.
Ultimately, I think you're going to have to work in internal coordinates
and eliminate the rotational and translational degrees of freedom from your
reaction coordinate definition. The best bet is to add restraints to your
dummy atom which fixes its position relative to the 'direction' you want to
pull along (you may need two distance restraints here).
As another thought, while you're using ParmEd and GB, you can modify the
exclusion list so any pairwise interactions between the dummy atom and the
protein are not calculated at all (although this is only a _very_ minor
performance benefit, since the omitted terms are all formally 0, anyway).
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 Tue Jan 22 2013 - 09:00:02 PST
