[gmx-users] PMF using umbrella sampling and Gromacs 5.0

[Christopher Neale]

Dear Eudes:

There are two issues: The first issue is the fact that you've got a sampling problem near the bilayer center. The second issue is the periodic bumps that you see in your PMFs. I'll take the second part first.

The source of the periodic bumps in PMFs from umbrella sampling is, to me, a million dollar question. I've seen them in my own work. I've seen then in the literature (as you noted for Fig. 4 in http://www.mdpi.com/1422-0067/13/11/14451/htm ). I've seen them when people use g_wham and Alan Grossfield's version of WHAM. What I don't yet know is if people also see this when running US simulations with AMBER, CHARMM, or NAMD. Frankly, either answer scares me a little. If you see wild oscillations of the PMF, at large distances then one possible source is that you are near a distance that is half your box length along the order parameter. Note that with constant pressure simulations you will get oscillations in the length of the box and also when your membrane changes shape it could flip the vector of closest approach by 180 deg if you are close to the half-box limit. However, that should not be relevant to the bumps (out of interest, how large is your box along z?). Although I don't know what is going on with these PMF bumps, I also note that membranes as a whole always tend to migrate toward positive z and lipids tend to flow to positive x in gromacs simulations, so I wonder if it is a rounding issue. Could you try again with double precision and see if you get the bumps?

As for the problem with sampling near the bilayer center... my first guess is that you've got some of your replicas on the wrong side of the bilayer's center. Did you intend to go across the center and sample also in the lower leaflet? I've only ever used gromacs 3 and 4 to do pulling simulations, never gmx 5. You can see my concerns about gmx5 for this type of issue near the bilayer center here: https://www.mail-archive.com/gromacs.org_gmx-users%40maillist.sys.kth.se/msg11324.html
One simply has to test it. However, if gmx5 does indeed work properly with this new US philosophy, then I presume that you simply need to move your starting coordinates such that your solute is in the positive leaflet for all replicas. You can test this by visualization or g_dist, which reports the sign of the displacement (look in the 5th column I think for the z value). Note that you are using pull_start=yes, so which side of the bilayer your solute is initially on should make a difference here.

Chris.
 
-- original message:

Hi Chris, a few days ago I posted a question on GMX list but unfortunately I not received an answer yet. So I write to you for help at your convenience.
http://permalink.gmane.org/gmane.science.biology.gromacs.user/78439

Besides the problem exposed the link above I'm trying to understand the importance of the relative distance between the COM. Due to fluctuation in the position of the center of mass position of the membrane, the minimum distance is never zero, as shown in the graphs of my pulling below. What is the effect of this difference in the calculation of the PMF? Could this be the source of my problem?
https://goo.gl/photos/cwwLzog6wXAgVJG88

Bests
eef

PS. I found this paper below where authors have published this unusual behavior (Figure 4). They did not explain the reason for the oscillations at large distances and I also do not know why this occurs.
http://www.mdpi.com/1422-0067/13/11/14451/htm



________________________________________
From: gromacs.org_gmx-users-bounces at maillist.sys.kth.se <gromacs.org_gmx-users-bounces at maillist.sys.kth.se> on behalf of Eudes Fileti <fileti at gmail.com>
Sent: 18 July 2015 21:06
To: gmx-users at gromacs.org
Subject: [gmx-users] PMF using umbrella sampling and Gromacs 5.0

Hello guys, I'm trying to calculate the PMF using umbrella sampling for a
small molecule to penetrate a lipid membrane. The 1D reaction coordinate is
along the z axis, which corresponds to the bilayer normal. The umbrella
potential acts on the center of mass of the molecule. The initial
configurations for each window, separated by a distance of 0.1nm, in a
range of 4 nm (distance between the centers of mass of the membrane and the
molecule), were obtained by a SMD.

The umbrella potential was applied according to the parameters below (for
Gromacs version 5.0.3). Each window was sampled by 5 ns.

; COM PULLING
pull                     = umbrella
pull_geometry            = direction
pull_dim                 = N N Y
pull_start               = yes
pull-print-reference     = no
pull_nstxout             = 1000
pull_nstfout             = 1000
pull_ngroups             = 2
pull-ncoords             = 1
; Group name, weight (default all 1), vector, init, rate (nm/ps),
kJ/(mol*nm^2)
pull-group1-name         = DPPC     ;  ref
pull-group2-name         = LIG         ;  pulled
pull-coord1-groups       = 1  2
pull-coord1-origin       = 0 0 0
pull-coord1-vec          = 0 0 1
pull-coord1-init         = 0
pull-coord1-rate         = 0.0
pull-coord1-k            = 3000

Histograms and the PMF obtained are shown in figure the link below.
https://goo.gl/photos/RkW9gbWrgeKEfV3R7

I repeated the same test using other parameters (larger and small k values)
and options (cylinder and position, for this later I have used gmx 4.6) and
so far I could not get a satisfactory profile. In all tests, a problem that
I have observed is that the region near z = 0 is not sampled (there is a
gap between 0.0 and 0.5 nm). In addition the profile presents not smooth,
but somehow oscillating mainly for large z.

As the gmx distance give a value different of zero for the smallest
distance of separation between the centers of mass (around 0.1-0.2 nm) I
believe that this weird behavior is related to the reference distance that
I have used.

Could someone give me a light? Most of the tips I read in GMX list are
related to previous versions of the Gromacs and in a way the other tips
were already included in my tests.

Thank you
eef

_______________________________________
Eudes Eterno Fileti
Instituto de Ciência e Tecnologia da UNIFESP
Rua Talim, 330, São José dos Campos - SP
Página: sites.google.com/site/fileti/
--
Gromacs Users mailing list

* Please search the archive at http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List before posting!

* Can't post? Read http://www.gromacs.org/Support/Mailing_Lists

* For (un)subscribe requests visit
https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users or send a mail to gmx-users-request at gromacs.org.