[gmx-users] Can g_wham support using different temperature for different windows?

Patrick Fuchs patrick.fuchs at univ-paris-diderot.fr
Wed Feb 23 12:04:05 CET 2011

I think your PMF is asymetric because your peptide is asymetric and you 
don't sample enough. To get a symetric PMF, your peptide would have to 
sample all the possible conformations *and* orientations in each window. 
Thus it means that for the windows in the center of the bilayer (where 
you say it's extended and interacts with the two monolayers) it'll have 
to rotate completely the other way round. This event will probably be 
*very* rare because you have to translocate positive charges across the 
membrane which cost ~ 40 to 50 kJ/mol (see the PMF of Lys+ and Arg+ in 
So as suggested by Chris, Justin and Xavier, you'll have to sample way 
more than 100 ns per window. I think you should go at least to the 
microsecond time scale (or more?). Or maybe starting from different 
initial conformations/orientations for a given window and then 
concatenate the different trajectories?
Also consider the remark of Xavier, TM or interfacial peptides are most 
of the time alpha-helical within the membrane. So far in literature, 
PMFs of a whole peptide across a bilayer were done by restraining the 
peptide in a helical conformation (e.g. 10.1016/j.bpj.2010.12.3682). It 
is anyway a very difficult problem (and probably impossible at atomistic 
resolution) to get a converged PMF for a whole peptide (e.g. 


Le 23/02/2011 05:25, Jianguo Li a écrit :
> Sorry, why do you think the PMF should be asymmetric?
> I pulled my peptide from d=9nm (above the membrane) to d=-3nm (below the
> membrane) and I did windowed umbrella sampling in the range of d=-1.05nm
> to d=9nm. At least the PMF should be symmetric with respect of the
> bilayer center in the range of d=[-1.05nm 1.05nm], something like a
> guassian distribution. But I got asymmetric PMF in this region. I also
> did reverse pulling starting from the peptide below the membrane ending
> with the peptide above the membrane. And the subsequent PMF of reversed
> pulling is also asymmetic.
> I have position restrains of the phosphate beads of the lipids in
> z-direction. So the membrane should be stable in REMD. But as you
> mentioned, if peptide is truly "stuck" in this orientation, REMD may be
> not useful. I will do a single simulation first at a higher temperature
> (e.g., 400K) in those bad windows to see if the peptide conformations
> are fully sampled.
> Cheers,
> Jianguo
> ------------------------------------------------------------------------
> *From:* Justin A. Lemkul <jalemkul at vt.edu>
> *To:* Gromacs Users' List <gmx-users at gromacs.org>
> *Sent:* Wednesday, 23 February 2011 10:24:46
> *Subject:* Re: [gmx-users] Can g_wham support using different
> temperature for different windows?
> Jianguo Li wrote:
>  > Thank you, Justin.
>  > Actually I did windowed umbrella simulations from d=-1.05nm to d=9nm.
> Since I think there is no problem in the region out of the membrane, so
> I only show the configurations within the membrane. My objective is to
> access the free energy barrier of the peptide translocate the negatively
> charged membrane. The problem is that the PMF is not symmetric with
> respect to the bilayer center due to the unconverged simulations.
> I would argue that the PMF is not symmetric because your reaction
> coordinate is not symmetric. How can you calculate a free energy of
> crossing a charged membrane when your peptide does not cross the
> membrane? What I proposed earlier was to obtain configurations at equal
> distances "above" and "below" the membrane (arbitrary in a periodic
> system, but hopefully you get the idea). If you can extract the peptide
> to the point where it is liberated from the membrane in the negative
> direction, I'd suspect you could solve your problem.
>  > Since g_wham does not support different temperatures in different
> windows, to increase the converges, I will probably consider to do REMD
> in those bad windows.
>  >
> This technique might work, provided you don't destabilize the membrane,
> but if the peptide is truly "stuck" in this orientation, I doubt that
> limited-range REMD would be very useful.
> -Justin
>  > Cheers
>  > Jianguo
>  >
>  > ------------------------------------------------------------------------
>  > *From:* Justin A. Lemkul <jalemkul at vt.edu <mailto:jalemkul at vt.edu>>
>  > *To:* Gromacs Users' List <gmx-users at gromacs.org
> <mailto:gmx-users at gromacs.org>>
>  > *Sent:* Tuesday, 22 February 2011 21:10:08
>  > *Subject:* Re: [gmx-users] Can g_wham support using different
> temperature for different windows?
>  >
>  >
>  >
>  > Jianguo Li wrote:
>  > > Thanks Justin and Chris and sorry for confusing interpretation.
>  > > Let me make it more clear. My peptide is flexible Martini beads,
> and highly positively charged. My membrane is a mixture of negatively
> charged lipids (25%) and zitterionic lipids(75%). So there is strong
> electrostatic attraction
>  > > between peptide and membrane. To get the PMF, I did the following:
>  > >
>  > > (1) I did pulling simulation along (0 0 -1) direction to pull my
> peptide across the membrane. Then I got different configurations
> corresponding to different windows along the reaction coordinates, which
> is the z-distance
>  > > between peptide and membrane. This figure
> (http://www.flickr.com/photos/lijg/5467080971/) shows some of the
> configurations at certain reaction coordinates.
>  > >
>  >
>  > Are you not sampling configurations outside of the membrane (i.e. in
> water)? I would think that would solve your problem. You don't show any
> configurations in which the peptide is completely dissociated from the
> membrane. I don't know your objectives, but I would think that if you
> could completely extract the peptide from the membrane after passing
> through it, this would solve your problem.
>  >
>  > > (2) In each window, I used the corresponding configuration that
> generated by the pulling simulation as initial input and run umbrella
> sampling. The size of each window is 0.15 nm, but close to the bilayer
> cneter (e.g., -0.6<d<0.6), I have
>  > > increased number of windows so that the width of the window is to
> be 0.05 or 0.1 nm, I also tried to use different force constant in these
> windows.
>  > >
>  > > From the figure (http://www.flickr.com/photos/lijg/5467080971/) ,
> we can classify the peptide conformation to be either extended
> (interacting with two bilayers) or compact (interacting with only one
> bilayer). Ideally, the peptide conformation should be similar for d=x
> and d=-x. The problem is that the configuration of peptide is not
> symmetric with respect to the bilayer center. For example, the peptide
> configuration is compact at d=0.6 and d=0.9, but the peptide is extended
> at d=-0.6 and d=-0.9. This leads Hysteresis. If I use g_wham to generate
> PMF, then the PMF is not symmetric with respect to the bilayer center.
> Using more number of windows and different force constant did not remove
> the problem.
>  > >
>  > > In my opinion, at least in some windows, the peptide should sample
> both compact and extended structure. But what I found is that the windowed
>  >
>  > Don't pre-judge the model :) Also, as I said before, there is no
> reason to suspect that MARTINI will produce any meaningful secondary
> structure changes. It was not parameterized to do so.
>  >
>  > > umbrella simulation depends on the initial peptide conformation. If
> the initial peptide conformation is compact, then after 100 ns, it is
> still compact; if the initial peptide in that window is extended, the
> final configuration is also extended. I also tried to run longer
> equilibrium time (e.g., 200 ns), but the problem still exists.
>  > >
>  >
>  > Sounds like a limitation of the force field model.
>  >
>  > > My question is how to increase sampling of the peptide
> conformation? I just think of two choices:
>  > > (1) use high temperature (e.g., 500K) at those bad windows. As I
> mentioned, I am wondering if g_wham can unbias the effect of using
> different temperatures in different windows.
>  > > (2) use REMD in those bad windows. These need a lot of
> computational resources.
>  > >
>  >
>  > Neither of these will be useful in generating a sensible PMF curve.
> WHAM needs a single temperature for proper weighting. If you start
> including different temperatures in different regions of phase space, I
> would imagine the weighting would be completely incorrect.
>  >
>  > Note that SMD is not the only option for generating starting
> configurations. If you think that certain orientations or configurations
> are "correct," you can build them yourself, but keep in mind that you'll
> have to justify this procedure to a skeptical audience.
>  >
>  > -Justin
>  >
>  > > Is there any other method to deal with the insufficient sampling?
>  > > Any suggestions are welcome, thanks for your time reading this email!
>  > >
>  > > Cheers,
>  > > Jianguo
>  > >
>  > >
>  > >
> ------------------------------------------------------------------------
>  > > *From:* Justin A. Lemkul <jalemkul at vt.edu <mailto:jalemkul at vt.edu>
> <mailto:jalemkul at vt.edu <mailto:jalemkul at vt.edu>>>
>  > > *To:* Gromacs Users' List <gmx-users at gromacs.org
> <mailto:gmx-users at gromacs.org> <mailto:gmx-users at gromacs.org
> <mailto:gmx-users at gromacs.org>>>
>  > > *Sent:* Tuesday, 22 February 2011 11:13:05
>  > > *Subject:* Re: [gmx-users] Can g_wham support using different
> temperature for different windows?
>  >

!!!! new E-mail address: patrick.fuchs at univ-paris-diderot.fr !!!!
Patrick FUCHS
Dynamique des Structures et Interactions des Macromolécules Biologiques
INTS, INSERM UMR-S665, Université Paris Diderot,
6 rue Alexandre Cabanel, 75015 Paris
Tel : +33 (0)1-44-49-30-57 - Fax : +33 (0)1-47-34-74-31
Web Site: http://www.dsimb.inserm.fr/~fuchs

More information about the gromacs.org_gmx-users mailing list