I do not think you will get meaningful results for diffusion constants. you
are using nscm, which removes part of the motion. your diffusion constants
are non-physical.
On Tue, Nov 23, 2010 at 6:04 PM, John S <s.john634.gmail.com> wrote:
> Thank You for your suggestion.I can use atomicfluct to get the deviation
> from the the average structure.
> Here I need to calculate the diffusion coefficients of selected residues
> too
> .
> So I plan to calculate this from the slope of Mean Square Deviation plots
> over the a selected part of trajectory (which is a range of 5000ps).
>
> I am not calculating diffusion of solvent , but of selected residues of my
> structure (in all 5 residues).
> Yes the MSD trajectory is noisy, will I get meaningful results out of this
> analysis of diffusion coefficients
>
> Thanks
> John
>
> On Tue, Nov 23, 2010 at 4:40 PM, Thomas Cheatham III <tec3.utah.edu>
> wrote:
>
> >
> > > To be more clear ,I am calculating as below,
> > >
> > > trjain md.x 4 40000 4
> > > strip :64-200
> > > diffusion :1 1 res1
> > > diffusion :2 1 res2
> >
> > If you are trying to calculate mean-squared-displacement (which is
> related
> > to the diffusion coefficient) as opposed to root-mean-squared deviation
> > (which is what Simmerling was referring to) you want to use the diffusion
> > command. You definately do not want to do any RMS fitting since this
> best
> > fits the coordinates via rotation/translation and will remove the net
> > displacement. There is also no need to strip any coordinates for this
> > calculation.
> >
> > Note that the MSD for a single residue will be rather noisy; to get
> > reasonable diffusion coefficients for a solvent it is much better to
> > average over all the solvent than you follow an individual solvent
> > molecule. Tracking an individual residue will likely, except if done as
> > an ensemble average over multiple trajectories, not provide much useful
> > information except to suggest that one residue moves "more" than another.
> >
> > Now maybe you do not want the mean-squared displacement but instead want
> > the atomic (or residue) positional fluctuations, such as a B-factor. For
> > that use the atomicfluct command and note that in this case, you likely
> > want to fit the trajectory to an average structure first to get the
> > fluctuations about the average structure...
> >
> > --tec3
> >
> > _______________________________________________
> > AMBER mailing list
> > AMBER.ambermd.org
> > http://lists.ambermd.org/mailman/listinfo/amber
> >
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Tue Nov 23 2010 - 16:00:03 PST
