[gmx-users] PMF steadily increasing

Mon Jun 20 17:56:32 CEST 2016

Good that they are not interacting at large restraint distances. If they have opposite net charges, then one expects some attraction at any displacement, no matter how large. Obviously this attraction should decay into the noise at some distance, but I don't know how far that is and it should depend on salt concentration and also for your system the distribution of charges. For instance, if the nearest ends of the peptide and protein are 2 nm apart and have opposite net charge it would not surprise me at all if there was some attraction. If you can convince yourself that this is what is going on I presume that you could do some type of fit and extrapolation of the PMF that you do have.

I am worried about your comment that you are selecting frames. If you mean to start the run, then fine, but if you're scripting the selection of distances (frames) to use as wham input, then you've gone down the wrong track. My point about wham is that if you tell your wham program that the umbrella is centered at 6.0 nm but it is really centered at 6.05 nm then your going to get the wrong PMF and if there is any systematic bias between real center and what you tell wham then you would get a slope to your PMF.

I suggest that you do some test runs with (a) two LJ spheres and (b) with a Na-Cl pair. The first will help to confirm that you have no other error in your method and the second will give you a feel for separation distances at which the PMF flattens (and for your macromolecule analogy then you have to think in terms of the locations of point charges, not make the analogy to the center of mass).

________________________________________
From: gromacs.org_gmx-users-bounces at maillist.sys.kth.se <gromacs.org_gmx-users-bounces at maillist.sys.kth.se> on behalf of Lukas Zimmermann <luk.zim91 at gmail.com>
Sent: 20 June 2016 06:03:51
To: gmx-users at gromacs.org
Subject: Re: [gmx-users] PMF steadily increasing

Thank you again for your remarks. This is what I found so far:

(1) With gmx mindist it was very clear that protein and peptide do not
interact. For the last 5 or 6 Umbrella Windows, the minimal distance
between the Pull groups was at least 2 nm
(2) That is indeed the case. The protein has net charge  -1 and the peptide
has charge +3.  Can you tell me what effect this might have?  I am unaware
that this might cause problems.
(3) I tried this, but this showed no apparent effect on the shape of the
curve.
(4) I do not understand in how far WHAM cares about the COM-COM distances.
I extract my frames from pull.xtc and I do not see each COM-COM distance
that would be required to have exactly, say, 0.1 nm
spacing, so I use a script which selects frames which resemble 0.1 nm
spacing as closely as possible, so there might some deviation.

2016-06-16 21:24 GMT+02:00 Lukas Zimmermann <luk.zim91 at gmail.com>:

> Thank you very much for your suggestions. I will check your individual
> remarks and provide feedback.
>
> 2016-06-15 19:01 GMT+02:00 Christopher Neale <chris.neale at alum.utoronto.ca
> >:
>
>> (1) Are the protein and peptide really never interacting at d=7 nm? I
>> presume you've got a peptide that would be maybe 5 nm long when fully
>> extended, and your dG minimum is at 1.5 nm, so giving half the peptide
>> length that would imply possible contact at 4 nm, so I expect 7 nm is
>> sufficient, but gmx mindist can tell you for certain. For example, if the
>> protein forms an overhanging pocket around the binding site, then it is
>> quite possible that an unfolded peptide would be sometimes (even rarely)
>> contacting a 568 residue protein even at a COM-COM distance of 7 nm.
>>
>> (2) net charge positive on one and negative on the other?
>>
>> (3) unconverged? Did you try eliminating the first half of your
>> production sampling and see if this fixes the issue?
>>
>> (4) did you do wham properly? Your mdp file indicate that your windows
>> are not *exactly* at 0.1 and 0.2 nm increments (use of
>> pull_coord1_start=yes), which is fine but only as long as wham doesn't
>> think your windows are exactly at these increments.
>>
>> There may be some entropy change as the peptide becomes unbound and can
>> then sample full X and Y, but on its own that should not continue to impact
>> the sampling after contact is permanently broken.
>> ________________________________________
>> From: gromacs.org_gmx-users-bounces at maillist.sys.kth.se <
>> gromacs.org_gmx-users-bounces at maillist.sys.kth.se> on behalf of Lukas
>> Zimmermann <luk.zim91 at gmail.com>
>> Sent: 15 June 2016 12:45:51
>> To: gmx-users at gromacs.org
>> Subject: [gmx-users] PMF steadily increasing
>>
>> Dear GROMACS community,
>>
>> I performed umbrella sampling study to estimate the binding free energy
>> between a globular
>> protein with 568 residues and a small peptide with 13 residues. I use the
>> pull code with k = 800 and rate = 0.005 to generate the initial
>> conformations over the time course of 1200 ps. The
>> center of masses distance between the pull groups ranges from  1.4 nm ad
>> 7.8 nm in the pull trajectory.
>>
>> I then extract conformations from pull.xtc with a spacing of 0.1 nm until
>> 3
>> nm distance
>> and a spacing of 0.2 nm for the remainder, yielding 38 windows in total.
>> After having
>> equilibrated each window with NVT and NPT under full position restraints,
>> I
>> conducted
>> production simulation under NPT ensemble for 14 ns for each window.
>> Finally, gmx wham
>> computed the PMF curve which you can see here:
>>
>> https://www.dropbox.com/s/fp7ol41qoyokmjm/profile.xvg?dl=0
>>
>> and this is the associated histogram:
>>
>> https://www.dropbox.com/s/bp6obymjc2qeu37/histo.xvg?dl=0
>>
>> Please find here my MDP pull parameters:
>>
>> pull                    = yes
>> pull_ngroups            = 2
>> pull_ncoords            = 1
>> pull_group1_name        = chainB    ; Protein
>> pull_group2_name        = chainC    ; Peptide
>> pull_coord1_type        = umbrella
>> pull_coord1_geometry    = distance
>> pull_coord1_groups      = 1 2
>> pull_coord1_dim         = N N Y
>> pull_coord1_rate        = 0.0
>> pull_coord1_k           = 800
>> pull_coord1_start       = yes
>>
>>
>> I would now be very interested why the curve does not become flat beyond
>> some certain distance, but rather seems to increase in a linear fashion,
>> though the distance between the pull groups is sufficiently large. Could
>> this be related to entropic effects? Is there a way to
>> have the PMF properly normalized?
>>
>> The force field here is GROMOS 53a6 with SPC water. Temperature is coupled
>> to
>> 310 K and pressure to 1 bar.  Cutoffs are 1.4 nm, longe range ES are
>> resolved with PME
>> and DispCorr is set to EnerPress. Bonds are constrained with LINCS.
>>
>> The Protein is prevented from rotation by having three CA atoms restrained
>> with FC 1000.
>>
>>
>> Thank you very much, I appreciate any help and suggestions.
>>
>> Lukas
>> --
>> 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.
>> --
>> 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.
>>
>
>
--
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.