Re: Umbrella Sampling plus restraints

From: Aron Broom (
Date: Wed Jul 18 2012 - 08:47:34 CDT

Hi Felipe,

If you apply this restraint, it will not change the manner in which you
generate the PMF, that is, you will still only use your reaction coordinate
forces and harmonic centers as input to WHAM with the data for the reaction
coordinate trajectory over that window.

This doesn't mean, however, that your restraint might not change the PMF.
The "real" PMF (at least as far as the forcefield is concerned), is the one
where there are no extra restraints, but, that assumes that you have
sampled long enough to get good coverage of all the relevant microstates.
So in your case with the base opening, the particular windows of the
reaction coordinate in which you see this happening might need to be
sampled for longer.

One problem I would say about your wanting to add the restraint is, what
reason do you have for believing the base DOESN'T open up? Do you trust
(to a reasonable extent) the forcefield parameters for the DNA? If so,
then the base must be opening up because there is a more favourable
interaction with the protein, and restraining it may give a false result on
the PMF where your binding appears weaker, or the activation energy higher,
than it really is.

Something you might try is 2D Umbrella sampling, in which the second
reaction coordinate is a distance or H-bonding or something for that
base-pair, then you could see if the lowest free energy path for separation
of your DNA-protein complex involves a movement of that base. If you don't
have a good reason for believing that base is entirely non-dynamic, then I
would suggest you might have found a rather interesting "gate-keeping"
element on your reaction path.

Anyway, something I would highly recommend is to use MetaDynamics initially
to look at this. It is far easier to setup than Umbrella Sampling,
requires less manual intervention (deciding on new force constants and
centers) and generally give a good rough idea of the PMF much faster, not
to mention that extending it to 2D (2 reaction coordinates) is trivial,
whereas Umbrella Sampling in 2D can be taxing. You can then use that
metadynamics bias in a static way and perform Umbrella Sampling on top of
it in order to refine the PMF and get error bars.


On Wed, Jul 18, 2012 at 8:18 AM, <>wrote:

> Dear all,
> We are working currently in some free energy calculations on a protein DNA
> complex using unbrella sampling with the minimal distance between the two
> molecules as the reaction coordinate. Everything goes OK except that the
> last base pair in the DNA opens from time to time and contacts the protein.
> We would like to restrain it in order to maintain the watson-crick pairing
> but we don't know if it would do any harm to the PMF. As we are using tcl
> forces we know exactly how much energy the bias is applying to the system
> so in principle the should be no mixing between the pulling restraint and
> the bais pairing restraint.
> We were then wondering if somebody can point to anything against this
> approach. Also, i am not sure now if we need to use only the bias energy
> for the unbrella restraint or it is better to add the ones from the bases
> restraint in order to do the histograming as well. In principle, i think
> you need to add all the extra energy added to obtain every particular
> microstate, but i am not sure.
> Any ideas?
> Regards
> Felipe

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

This archive was generated by hypermail 2.1.6 : Tue Dec 31 2013 - 23:22:16 CST