[gmx-users] RDF(PMF) and Umbrella sampling

Gavin Melaugh gmelaugh01 at qub.ac.uk
Tue Jan 10 11:47:43 CET 2012


Hi Justin

Again, many thanks for the reply.
So when the COM distance changes sign, what effect does that have on the
distribution of the COM distance about the mean value for that window
i.e. If say my ref dist in 0 nm and the umbrella sampling allows the
distance to sample distances say at 0.02 nm to -0.02nm. What happens to
negative values? Obviously they are not counted as negative in the
distribution or else it would be centred at zero/

Cheers

Gavin

Justin A. Lemkul wrote:
>
>
> Gavin Melaugh wrote:
>> Hi Justin
>>
>> Thanks very much. One last question. What do you mean when you say "COM
>> reference distance is changing signs"? I thought  the COM distance was
>> the absolute distance between the two groups and therefore cannot be
>> negative?
>>
>
> The pull code deals in vectors.  Signs can change.  The use of
> "distance" as a geometry is perhaps somewhat misleading.
>
> -Justin
>
>> Cheers
>>
>> Gavin
>>
>> Dariush Mohammadyani wrote:
>>> Hi Gavin,
>>>
>>> A question arose for me: why did you consider the (rate = 0)?
>>>
>>> Dariush
>>>  
>>>
>>> On Fri, Jan 6, 2012 at 11:47 AM, Gavin Melaugh <gmelaugh01 at qub.ac.uk
>>> <mailto:gmelaugh01 at qub.ac.uk>> wrote:
>>>
>>>     Hi Justin
>>>
>>>     Just a quick clarification regarding my previous point. With
>>>     geometry =
>>>     distance, and pull_dim =Y Y Y . Is the pull_group sampling all
>>>     dimensions equally (or without prejudice) about pull_init ?  And
>>>     iN your
>>>     first reply what did you mean about by "straight pull" ?
>>>
>>>     Cheers
>>>
>>>     Gavin
>>>
>>>     Justin A. Lemkul wrote:
>>>     >
>>>     >
>>>     > Gavin Melaugh wrote:
>>>     >> Hi Justin
>>>     >>
>>>     >> Thanks for the reply. I wanted my "pulling" to be free in all
>>>     >> directions, that is in the liquid state with no defined reaction
>>>     >> coordinate i.e not along a specific axis. This is why I used
>>>     geometry =
>>>     >> distance. Would you agree with this approach?
>>>     >
>>>     > I suppose there is an argument that can be made for a more free
>>>     > approach such as this one, but you're going to get the
>>> artifact you
>>>     > observed the instant your pull group moves past a zero COM
>>> distance.
>>>     > Whether or not this is a significant problem is something you'll
>>>     have
>>>     > to determine.
>>>     >
>>>     > -Justin
>>>     >
>>>     >> By free I mean. The absolute distance between the COG of the
>>>     ref group
>>>     >> and that of the pull group.
>>>     >>
>>>     >> Cheers
>>>     >>
>>>     >> Gavin
>>>     >>
>>>     >> Justin A. Lemkul wrote:
>>>     >>>
>>>     >>> Gavin Melaugh wrote:
>>>     >>>> Dear all
>>>     >>>>
>>>     >>>> I have a query regarding umbrella sampling simulations that I
>>>     have
>>>     >>>> carried out to study a dynamical process of a guest inserting
>>>     into a
>>>     >>>> host. I always get get a wall tending off to infinity at or
>>> just
>>>     >>>> before
>>>     >>>> the zero distance between the
>>>     >>>> two species.
>>>     >>>> The process I describe, for one system in particular, happens
>>>     readily
>>>     >>>> and I have compared the PMF from a non constrained simulation
>>>     (via the
>>>     >>>> RDF and reversible work theorem) and the same PMF from a
>>> set of
>>>     >>>> umbrella sampling
>>>     >>>> simulations. They agree quite well but in the non constrained
>>>     >>>> simulation
>>>     >>>> I get a minimum practically at zero whereas for the umbrella
>>>     sampling
>>>     >>>> the minimum is shifted and there is an infinite wall close to
>>>     zero.
>>>     >>>> This
>>>     >>>> wall is not present from the reversible work theorem. Why the
>>>     infinite
>>>     >>>> wall? Why does the black histogram not centre around zero. Is
>>>     this an
>>>     >>>> artefact of the umbrella technique? Please see attached the
>>>     profile
>>>     >>>> from
>>>     >>>> the umbrella sampling technique, and the corresponding
>>>     histograms.
>>>     >>>>
>>>     >>> What's happening is the COM reference distance is changing
>>>     signs, so
>>>     >>> you get an artifact.  The "distance" geometry is relatively
>>>     inflexible
>>>     >>> and is only suitable for straight pulls of continuously
>>>     increasing or
>>>     >>> continuously decreasing COM distance.  You should try using the
>>>     >>> "position" geometry instead.  There are some notes that you
>>>     may find
>>>     >>> useful in my tutorial:
>>>     >>>
>>>     >>>
>>>    
>>> http://www.bevanlab.biochem.vt.edu/Pages/Personal/justin/gmx-tutorials/umbrella/05a_pull_tips.html
>>>
>>>     >>>
>>>     >>>
>>>     >>>
>>>     >>> -Justin
>>>     >>>
>>>     >>>> Here is an excerpt from one of the umbrella mdp files.
>>>     >>>>
>>>     >>>> pull        = umbrella
>>>     >>>> pull_geometry = distance
>>>     >>>> pull_dim = Y Y Y
>>>     >>>> pull_start = no
>>>     >>>> pull_ngroups = 1
>>>     >>>> pull_group0 = cage_1
>>>     >>>> pull_group1 = tail
>>>     >>>> pull_init1 = 0
>>>     >>>> pull_rate1 = 0.0
>>>     >>>> pull_k1 = 10000
>>>     >>>> pull_nstxout = 150
>>>     >>>> pull_nstfout = 150
>>>     >>>>
>>>     >>>>
>>>     >>>> Cheers
>>>     >>>>
>>>     >>>> Gavin
>>>     >>>>
>>>     >>>>
>>>     >>>>
>>>    
>>> ------------------------------------------------------------------------
>>>
>>>     >>>>
>>>     >>>>
>>>     >>>>
>>>     >>>>
>>>    
>>> ------------------------------------------------------------------------
>>>
>>>     >>>>
>>>     >>>>
>>>     >>
>>>     >>
>>>     >
>>>
>>>     --
>>>     gmx-users mailing list    gmx-users at gromacs.org
>>>     <mailto:gmx-users at gromacs.org>
>>>     http://lists.gromacs.org/mailman/listinfo/gmx-users
>>>     Please search the archive at
>>>     http://www.gromacs.org/Support/Mailing_Lists/Search before posting!
>>>     Please don't post (un)subscribe requests to the list. Use the
>>>     www interface or send it to gmx-users-request at gromacs.org
>>>     <mailto:gmx-users-request at gromacs.org>.
>>>     Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>>>
>>>
>>>
>>>
>>> -- 
>>> Kind Regards,
>>> Dariush Mohammadyani
>>> Department of Structural Biology
>>> University of Pittsburgh School of Medicine
>>> Biomedical Science Tower 3
>>> 3501 Fifth Avenue
>>> Pittsburgh, PA 15261
>>> USA
>>>
>>
>




More information about the gromacs.org_gmx-users mailing list