Re: Metadynamics and Colvar Width

From: Aron Broom (broomsday_at_gmail.com)
Date: Mon Oct 29 2012 - 12:53:22 CDT

You are correct that the width controls the width of the hills.

I smaller width will yield a more detailed (i.e. higher resolution) PMF,
but because the hills are narrower, it will take longer to fill in that
PMF.

I think it many cases you can choose the width based on an intuitive sense
of the colvar under investigation, and what kind of information you want.
For instance, if you are looking at ligand binding, a width of 1 angstrom
is probably fairly appropriate, since you are unlikely to be interested in
the noisy detailed information below that, but it is still narrow enough to
capture the relevant movements that will be important for accuracy.

Another approach you can use, rather than trying to determine what width
you would want, is to think more about how many "bins" are appropriate
along your reaction coordinate. I think some work by Parinello (one of the
researchers who developed metadynamics) showed that you can get good
information with as few as ~10 bins along your reaction coordinate. So in
the above case, if the relevant range for your ligand binding to the
binding site is 0 - 20 angstroms, you could set the width to 2 angstroms.

Overall it's a bit subjective I suppose, but it's better to start with
wider bins, and optimize from there, rather than waste your time getting
unnecessarily high resolution PMFs (especially since PMFs may need a lot of
sampling time as it is in order to converge properly).

~Aron

On Mon, Oct 29, 2012 at 10:19 AM, Marc Gordon <marcgrdn55_at_gmail.com> wrote:

> Hello all
>
> I'm performing some metadynamics investigations and I am confused as to
> what colvar width is doing exactly.
>
> After doing some reading in the user guide it seems that width is involved
> in determining the width of new hills but I'm not sure how I would go about
> finding an appropriate value for width. Is anyone with experience with this
> sort of thing able to give me some advice or perhaps point me to some
> material I could look at to help me with this?
>
>
> Thanks
>

-- 
Aron Broom M.Sc
PhD Student
Department of Chemistry
University of Waterloo

This archive was generated by hypermail 2.1.6 : Mon Dec 31 2012 - 23:22:12 CST