From: Aron Broom (broomsday_at_gmail.com)
Date: Mon Feb 03 2014 - 15:01:34 CST
That sounds mostly right. Maybe a few points of confusion:
1) You don't necessarily need to pre-generate all the coordinates. You can
start from a single structure and the restraint you apply should get you
where you want to be. Keep in mind that because the energy landscape isn't
flat, whatever angle you are looking at won't have a mean at the value of
2) This is good. You may not need to minimize. It depends on how strong
the restraints are, and how far from the restraint value you are starting.
Minimizing is probably fairly safe, but check the final structures from
the minimization to ensure nothing got mangled.
3) This is the confusing line. I'm not sure what you mean about a specific
energy in a direction. Once you have the harmonic restraint you applied in
2, you have all you need. You do of course need to choose the equilibrium
value and force constant for that restraint, but you'd do that in 2. So
all you need to do here is run the simulation for some amount of time.
4) You don't need to compile the trajectories together. All you need to do
is setup the colvars in NAMD to output the value of your restraint, and
then use WHAM on those lists of values. Each window should have its own
file and it's own force constant and equilibrium value.
There are two main stumbling points with Umbrella Sampling, one at the
beginning and one at the end:
In the beginning, the problem you'll face is choosing the right force
constants and equilibrium values for your restraints. In theory there are
no wrong choices, as any choice can help define your PMF, but practically,
you want the fewest number of windows that overlap with adjacent windows to
span your reaction coordinate. Regions on the PMF where the slope is steep
will be hard to explore, because the coordinate will tend to move down that
slope. Not all windows need to use the same force constant, and the
spacing between windows does not need to be even.
At the end, the resulting PMF is only meaningful (accurate?) if the
simulations in ALL the windows have reached equilibrium. That is, have all
settled into whatever energy minimum is created by the combination of the
native landscape and the restraint. It's generally very hard to know when
this has happened. The easiest solution on a broad scale is to do your
umbrella sampling twice, with different starting coordinates, and different
langevin seeds for the thermostat. Then, once you've run for some amount
of time that you hope will be good, compile both PMFs and see if they are
equivalent. If they are not, you need to run for longer. Obviously, this
also means you need to be trimming off the first part of each window's
trajectory, since the assumption is that part isn't at an equilibrium value.
On Mon, Feb 3, 2014 at 2:40 PM, Dan F. X. Martingano <Dynamic00_at_msn.com>wrote:
> > On Feb 3, 2014, at 2:36 PM, "Aaron Larsen" <
> alarsen_at_molbio.mgh.harvard.edu> wrote:
> > Hello,
> > I have recently performed equilibrium calculations on short
> oligonucleotides to investigate factors influencing the pseudorotation of
> the sugar pucker. I would like to extend this work to include Umbrella
> sampling to establish the energy landscape along this coordinate.
> > There have been some similar questions asked here but none seem to apply
> very directly to my situation.
> > First, I would like to establish that I have the correct notion as to
> the steps involved in this process.
> > 1) Generate a series of coordinates for a given system along the chosen
> reaction coordinate. In my case, this would be a nucleotide with different
> values, P, for the sugar pucker.
> > 2) For each set of coordinates, apply a restraint and then minimize the
> > 3) For each system. apply a specific energy in a particular direction
> and record the resulting trajectories.
> > 4) Compile the trajectory files and use WHAM to investigate the PMF
> along the chosen coordinate.
> > Assuming that this is correct, is there any tutorial that can get me
> started on this process? I was unable to find reference to Umbrella
> sampling in the tutorial on the webpage but it might be covered by a
> different term.
> > Best,
> > Aaron
-- Aron Broom M.Sc PhD Student Department of Chemistry University of Waterloo
This archive was generated by hypermail 2.1.6 : Wed Dec 31 2014 - 23:22:04 CST