[gmx-users] wham question

[David Mobley]

Manohar,

I don't have a tutorial, but have done a bit of this. I think GROMACS
has some built in tools for WHAM, but I have never used these. I just
apply harmonic restraints to pull a ligand out of the binding site and
then use my own WHAM code to analyze the data.

In terms of picking the number of umbrellas and the spring constants,
there are semi-automated or adaptive ways to do this, but none that
are implemented in GROMACS, I think. So my approach so far has been to
specify a set of umbrella centers and spring constants, then run some
fairly short (i.e. 100 ps) trajectories and look at the resulting
distance distributions. You want to make sure you get good overlap.
Usually it takes some iteration before you get to the point where you
have enough umbrellas, centered at the right places, with the right
spring constants, in order to ensure good overlap. This, however, is
key -- without good overlap, you'll get an answer, but it will be
wrong (or, if you're lucky/unlucky, depending on your perspective,
right for the wrong reason).

After I find something that gives me reasonable-looking overlap for
short trajectories, then I go ahead and run longer "production"
simulations before doing the analysis.

The additional complication is how exactly you pull the ligand out of
a binding site. You might want to look at the PNAS paper by Woo and
Roux (102:6825-6830) for an example. One issue is that if you just
pull the ligand with simple distance restraints, there is no guarantee
that you will pull it in a sane direction. This also may be especially
problematic when the ligand begins leaving the protein, as at any
fixed distance from a reference atom in the protein, it will be able
to sample everywhere on a spherical shell, which can take a long time
to equilibrate.

My approach so far has been to try and use cartesian restraints to
pull the ligand out along a particular axis that I specify in advance,
and then compute the free energy of turning off the cartesian
restraints at the endpoints. I'm not sure that this will work, yet,
but at least it gets around the problems I just described.

I'm not sure how Woo and Roux got around this; it sounds like they
pulled along a particular axis; perhaps the ability to do that is
implemented in CHARMM. You might be able to also do this by using an
appropriate combination of distance and angle restraints between some
atoms in the protein and ligand, in order to specify a direction.

To make a long story short: I don't have a silver bullet, sorry.

David


On 9/12/06, Manohar Murthi <sir_donald_bradman at yahoo.com> wrote:
> hi all
>
> does anyone have an example/tutorial of how to do
> umbrella sampling using gromacs & wham?
>
> i'm about to try to calculate free energies of binding
> using this method.
>
> i can find no information whatsoever about 'wham.gct',
> which is an optional input mentioned in the mdrun
> online reference page. what is this file supposed to
> contain?
>
> i've yet to actually try it, but setting up the other
> input files seems relatively straightforward.
>
> i'm hoping to benefit from others' experience to
> figure out how to choose and vary the harmonic force
> constants for the different umbrella runs.
>
> any advice would be greatly appreciated.
>
> cheers
> mo
>
>
>
> __________________________________________________
> Do You Yahoo!?
> Tired of spam?  Yahoo! Mail has the best spam protection around
> http://mail.yahoo.com
> _______________________________________________
> gmx-users mailing list    gmx-users at gromacs.org
> http://www.gromacs.org/mailman/listinfo/gmx-users
> Please don't post (un)subscribe requests to the list. Use the
> www interface or send it to gmx-users-request at gromacs.org.
> Can't post? Read http://www.gromacs.org/mailing_lists/users.php
>