[gmx-users] gmx wham (again)

[Jochen Hub]

Hi Alex, hi Justin,

Am 18.04.18 um 09:17 schrieb Alex:
> I suppose this question is mostly for Justin...
> 
> Let me remind what I am dealing with and ask if my idea is correct.
> 
> I have a rectangular membrane in XY with a pore at (X/2, Y/2) in water 
> and want to get the Gibbs free energy curve for an ion. For this, I have 
> a bunch of starting configurations at (X/2, Y/2) and Z varying between 
> some -z0 and z0. The bias in the "fake" pull mdp is applied as N N Y. 
> Near the membrane, this means the entire plane is sampled, which adds 
> contributions I am not interested in. I want the pore and a small region 
> of the membrane around it, not the membrane, given its propertie
> 
> So, I applied weak (k=50) in-plane restraint to the ion for each of the 
> sampled configurations -- to keep the sampling region a bit closer to 
> the pore. The results look completely different, but they finally make 
> very good qualitative sense. The ion still walks around within a small 
> disk, but not much -- this tentatively makes me happy.

What you want is a /cylindrical flat-bottomed (FB) position restraint/ 
around the pore, see the manual for "flat-bottomed position restraints". 
You will make sure that the ion stays close the pore in the xy-plane, 
and this way you avoid sampling problems at the pore entry and exit.

Note that the FB position restraint will change the entropy of the ion 
in the x-y plane. The entropic correction needed for your PMFs (to 
remove the contribution form the flat-bottomed restraint) is worked out 
in Appendix of this paper:

http://dx.doi.org/10.1021/acs.jpcb.6b11279

This correction allows you to refer your PMF to a well-defined /area per 
pore/ or, equivalently, /density of pores/ in the membrane (which 
sometimes causes a lot of confusion).

I would definitely not use a simple (not flat-bottomed) position 
restraint to restrain the ion near the pore - this may cause artifacts 
in your PMF.

The cylinder-based reaction coordinate (mentioned by Justin) affects how 
the center of mass of the membrane is computed in Z-direction. Hence, it 
changes the *reaction coordinate*. But it does *not* affect the sampled 
xy-plane of the ion.

The idea behind the cylinder-based reaction coordinate is to avoid that 
permeation barriers are smeared out due to membrane undulations. Hence, 
for a small membrane that hardly undulates, a simple reaction coordinate 
is probably fine (using pull_coord1_geometry = direction or 
direction-periodic).

I hope this helped.

Cheers,
Jochen

> 
> Would you believe the results obtained this way, assuming otherwise 
> proper setup?
> 
> Thanks,
> 
> Alex
> 

-- 
---------------------------------------------------
Dr. Jochen Hub
Computational Molecular Biophysics Group
Institute for Microbiology and Genetics
Georg-August-University of Göttingen
Justus-von-Liebig-Weg 11, 37077 Göttingen, Germany.
Phone: +49-551-39-14189
http://cmb.bio.uni-goettingen.de/
---------------------------------------------------