[gmx-users] Re: umbrella sampling with "pull=constraint"

Justin A. Lemkul jalemkul at vt.edu
Mon Nov 7 18:38:58 CET 2011



Vijayaraj wrote:
> My system is a self-assembled cyclic peptide, I am trying to extract PMF 
> to pull one of the terminal cyclic peptide. I have collected 25 windows 
> starting from 0.5nm with 0.05nm window size. the histogram shows up to 
> 0.7nm all the windows are overlapped around 0.5nm, and poor sampling 
> from 0.6 to 0.75nm. from 0.75nm the sampling is perfect. each cyclic 
> peptide has 8 residues and the terminal one forms 8 hbonds with the 2nd 
> cyclic peptide. I guess due to this strong hbonding, the sampling window 
> up to 0.7nm is very poor. If we extend the simulation only for these 
> windows can we get proper sampling windows? this observation is from 
> last 10ns of 20ns simulation.
> 

It sounds like you need a substantially stronger force constant to restrain the 
positions where the sampling is failing.

-Justin

> 
>     Message: 3
>     Date: Mon, 07 Nov 2011 10:20:07 -0500
>     From: "Justin A. Lemkul" <jalemkul at vt.edu <mailto:jalemkul at vt.edu>>
>     Subject: Re: [gmx-users] Re: umbrella sampling with "pull=constraint"
>     To: Discussion list for GROMACS users <gmx-users at gromacs.org
>     <mailto:gmx-users at gromacs.org>>
>     Message-ID: <4EB7F727.3000504 at vt.edu <mailto:4EB7F727.3000504 at vt.edu>>
>     Content-Type: text/plain; charset=ISO-8859-1; format=flowed
> 
> 
> 
>     Vijayaraj wrote:
>      > and also I tried to plot the PMF curve from this data, I got few
>     warning
>      > messages like "no data point in bin 20, You may not get a reasonable
>      > profile. Check your histograms!". the histogram shows poor
>     sampling from
>      > 0.6 nm to 0.7 nm, and the windows in this position completely
>     overlaps
>      > at 0.5-0.6 nm. should we have to extend the simulation for 0.6-0.7 nm
>      > window to get the restrain distance converged?
>      >
> 
>     It is still somewhat difficult to provide any useful advice.  You
>     haven't stated
>     what your system is, how your windows were assembled, at what
>     interval they
>     exist, or how many you have.  The above warnings suggest that you've
>     either set
>     up the windows wrong, have insufficient data, or are analyzing parts
>     of the
>     trajectory you shouldn't be (i.e. initial times that are equilibration).
> 
>     -Justin
> 
>      > On Mon, Nov 7, 2011 at 3:53 PM, Vijayaraj <vijayaraj81 at gmail.com
>     <mailto:vijayaraj81 at gmail.com>
>      > <mailto:vijayaraj81 at gmail.com <mailto:vijayaraj81 at gmail.com>>> wrote:
>      >
>      >     Hello,
>      >
>      >     Thanks for the suggestions. I did umbrella simulation for
>     20ns and
>      >     the pull force is as below.
>      >
>      >
>      >     19960.0000    -6.78192
>      >     19962.0000    33.3579
>      >     19964.0000    -3.1808
>      >     19966.0000    -15.0304
>      >     19968.0000    -55.4436
>      >     19970.0000    -38.9422
>      >     19972.0000    -9.41927
>      >     19974.0000    -5.95981
>      >     19976.0000    0.626319
>      >     19978.0000    2.28736
>      >     19980.0000    23.5115
>      >     19982.0000    10.6415
>      >     19984.0000    -0.233446
>      >     19986.0000    8.26525
>      >     19988.0000    36.7656
>      >     19990.0000    26.6017
>      >     19992.0000    64.43
>      >     19994.0000    53.3844
>      >     19996.0000    18.3834
>      >     19998.0000    39.8988
>      >     20000.0000    43.6291
>      >
>      >     I expect the force should be converged. my understanding might be
>      >     wrong. I have poor knowledge about this umbrella sampling
>      >     convergence and image selection for the PMF calculation. is this
>      >     fluctuation in force due to pulling group motion out of box?
>     can you
>      >     give me some idea about the selection of images for PMF
>     calculation?
>      >     I found from the pullx files that the COM distance converges
>     after
>      >     few ns of simulation. I understood that we cant do PMF study with
>      >     pull=constraint.
>      >
>      >     Vijay.
>      >
>      >
>      >         Message: 2
>      >         Date: Mon, 07 Nov 2011 09:07:21 -0500
>      >         From: chris.neale at utoronto.ca
>     <mailto:chris.neale at utoronto.ca> <mailto:chris.neale at utoronto.ca
>     <mailto:chris.neale at utoronto.ca>>
>      >         Subject: [gmx-users] umbrella sampling with "pull=constraint"
>      >         To: gmx-users at gromacs.org <mailto:gmx-users at gromacs.org>
>     <mailto:gmx-users at gromacs.org <mailto:gmx-users at gromacs.org>>
>      >         Message-ID:
>     <20111107090721.jy8zfzhvh4ccw0ws at webmail.utoronto.ca
>     <mailto:20111107090721.jy8zfzhvh4ccw0ws at webmail.utoronto.ca>
>      >        
>     <mailto:20111107090721.jy8zfzhvh4ccw0ws at webmail.utoronto.ca
>     <mailto:20111107090721.jy8zfzhvh4ccw0ws at webmail.utoronto.ca>>>
>      >         Content-Type: text/plain;       charset=ISO-8859-1;    
>     DelSp="Yes";
>      >                format="flowed"
>      >
>      >         Dear Vijay:
>      >
>      >         Can you please provide evidence for your claim that the
>     harmonic
>      >         potential is not applied properly, since you may decide
>     to use
>      >         pull=umbrella once you have set that up correctly.
>     Importantly,
>      >         movement out of the unit-cell is not a problem, as
>     discussed a
>      >         lot on
>      >         this list. Nevertheless, you do need to worry about which
>     images are
>      >         used to derive the pulling forces. You can often do that by a
>      >         judicious selection of pull_pbcatom (read about that
>     on-list and in
>      >         the manual). In some cases, however, pull_pbcatom can not
>     assist
>      >         enough and you are forced to make the box larger. None of
>     this is
>      >         alleviated by pull=constraint, which is why I'm not going
>     to answer
>      >         your question about your .mdp parameters at this point. Let's
>      >         get the
>      >         problems identified and solved first with pull=umbrella.
>      >
>      >         Chris.
>      >
>      >
>      >         -- original message --
>      >
>      >         Hello,
>      >
>      >         I have done umbrella sampling with "pull=umbrella" and I
>     found
>      >         that the
>      >         pulling group has high fluctuations and sometimes moving
>     out of the
>      >         periodic box. I think that the harmonic potential is not
>      >         properly applied
>      >         and thus the pulling group is not retained with the specified
>      >         COM distance
>      >         between the reference and pulling group. Can we use
>      >         pull=constraint option
>      >         to retain the pulling group within the COM distance? my
>     pulling
>      >         code is as
>      >         bellow with restrain at 0.5nm distance from ref group. I just
>      >         want to get
>      >         some idea about this pull code modification.
>      >
>      >         pull            = constraint
>      >         pull_geometry    = distance
>      >         pull_dim        = N N Y
>      >         pull_start      = no
>      >         pull_ngroups    = 1
>      >         pull_group0     = Chain_A
>      >         pull_group1     = Chain_B
>      >         pull_init1      = 0.50
>      >         pull_rate1      = 0.0
>      >         pull_k1         = 1000
>      >         pull_nstxout    = 500
>      >         pull_nstfout    = 500
>      >
>      >
>      >         Regards,
>      >         Vijay.
>      >
>      >
>      >
>      >         -
>      >
>      >
>      >
>      >
>      >
> 
>     --
>     ========================================
> 
>     Justin A. Lemkul
>     Ph.D. Candidate
>     ICTAS Doctoral Scholar
>     MILES-IGERT Trainee
>     Department of Biochemistry
>     Virginia Tech
>     Blacksburg, VA
>     jalemkul[at]vt.edu <http://vt.edu> | (540) 231-9080
>     http://www.bevanlab.biochem.vt.edu/Pages/Personal/justin
> 
>     ========================================
> 
> 

-- 
========================================

Justin A. Lemkul
Ph.D. Candidate
ICTAS Doctoral Scholar
MILES-IGERT Trainee
Department of Biochemistry
Virginia Tech
Blacksburg, VA
jalemkul[at]vt.edu | (540) 231-9080
http://www.bevanlab.biochem.vt.edu/Pages/Personal/justin

========================================



More information about the gromacs.org_gmx-users mailing list