[gmx-users] Monte Carlo with Gromacs and implicit solvent

Pool, R. r.pool at vu.nl
Wed Feb 1 09:08:22 CET 2012


Hi Gianluca,

I have no experience with the -rerun option of mdrun, but my guess is that an integrated MC/NVT algorithm in GromPy will be faster.

For CGMC, a trial configuration involves generating a new tpr file using grompp, either with N+1 or N-1 molecules of interest. I implemented such a shell approach to compare it to GromPy. It turns out that GromPy is 14 times faster compared to the shell approach for system sizes of about 500 molecules. This speed-up factor increases with increasing system size. This all is due to the fact that tpr's are kept in memory during run time in GromPy. The shell approach on the other hand uses up a lot of computer time for e.g. file I/O.

The above is based on the GCMC case so it would not necessarily have to apply to MC/NVT.

Cheers,
Rene
=====================================================
René Pool

Division of Molecular and Computational Toxicology
Department of Chemistry and Pharmaceutical Sciences
Vrije Universiteit Amsterdam
De Boelelaan 1083
1081HV AMSTERDAM, the Netherlands
-----
IBIVU/Bioinformatics
Department of Computer Science
Vrije Universiteit Amsterdam
De Boelelaan 1081a
1081HV AMSTERDAM, the Netherlands

Room P 2.75
E: r.pool at vu.nl
T: +31 20 598 76 12
F: +31 20 598 76 10
=====================================================

________________________________________
From: Gianluca Interlandi [gianluca at u.washington.edu]
Sent: 31 January 2012 19:55
To: Pool, R.
Cc: Discussion list for GROMACS users
Subject: RE: [gmx-users] Monte Carlo with Gromacs and implicit solvent

Hi Rene,

Thanks for your reply. One question though. How much overhead do you think
I would have in calling "mdrun -rerun" after every step and would it run
more efficiently if I used GromPy?

Thanks,

      Gianluca

On Tue, 31 Jan 2012, Pool, R. wrote:

> Hi Gianluca,
>
> Ah, I get it. In that case you can actually use GromPy, but it needs to be modified slightly. You need only one tpr file that will be stored in memory during execution of GromPy. For the sampling you can generate new configurations and accept/reject them according to the appropriate rules. The coordinates/velocities have to be put back in the "tpr" data structure before calculating the energy. You have to write a routine that generates a new configuration yourself, but that should not be too hard. Obviously, the Monte Carlo acceptance rules have to be adjusted for sampling in the NVT ensemble.
> The alternative is mdrun -rerun as suggested before.
>
> Cheers,
> René
> =====================================================
> René Pool
>
> Division of Molecular and Computational Toxicology
> Department of Chemistry and Pharmaceutical Sciences
> Vrije Universiteit Amsterdam
> De Boelelaan 1083
> 1081HV AMSTERDAM, the Netherlands
> -----
> IBIVU/Bioinformatics
> Department of Computer Science
> Vrije Universiteit Amsterdam
> De Boelelaan 1081a
> 1081HV AMSTERDAM, the Netherlands
>
> Room P 2.75
> E: r.pool at vu.nl
> T: +31 20 598 76 12
> F: +31 20 598 76 10
> =====================================================
>
> ________________________________________
> From: Gianluca Interlandi [gianluca at u.washington.edu]
> Sent: 30 January 2012 17:36
> To: Pool, R.
> Cc: Discussion list for GROMACS users
> Subject: RE: [gmx-users] Monte Carlo with Gromacs and implicit solvent
>
> Hi René,
>
> I want to perform simulations of a protein near a surface, for example
> silicates. The surface is fixed and the protein is treated as a rigid
> body. Solvation is treated with GBSA implicit model. The point is to
> sample orientations of the protein with respect to the surface using Monte
> Carlo moves. I guess I am sampling the canonical ensemble.
>
> Gianluca
>
> On Mon, 30 Jan 2012, Pool, R. wrote:
>
>> Hi Gianluca,
>>
>> The tutorial involves simulations using a simple LJ model. GromPy/GCMC has been validated by comparing equations of state, calculated using MD in the NVT ensemble and GromPy in the muVT ensemble, for both the LJ and SPC fluids. For both models, the muVT and NVT results are in agreement. Therefore, in principle, you can use GromPy for more sophisticated molecular models.
>>
>> Which ensemble are you intending to use?
>>
>> Cheers,
>> René
>> =====================================================
>> René Pool
>>
>> Division of Molecular and Computational Toxicology
>> Department of Chemistry and Pharmaceutical Sciences
>> Vrije Universiteit Amsterdam
>> De Boelelaan 1083
>> 1081HV AMSTERDAM, the Netherlands
>> -----
>> IBIVU/Bioinformatics
>> Department of Computer Science
>> Vrije Universiteit Amsterdam
>> De Boelelaan 1081a
>> 1081HV AMSTERDAM, the Netherlands
>>
>> Room P 2.75
>> E: r.pool at vu.nl
>> T: +31 20 598 76 12
>> F: +31 20 598 76 10
>> =====================================================
>>
>> ________________________________________
>> From: Gianluca Interlandi [gianluca at u.washington.edu]
>> Sent: 24 January 2012 21:20
>> To: Pool, R.; Discussion list for GROMACS users
>> Subject: Re: [gmx-users] Monte Carlo with Gromacs and implicit solvent
>>
>> Hi Renè,
>>
>> This might be a silly question. In the documentation of GromPy it says
>> that:
>>
>> "The system consists of mono-atomic water molceules defined in the MARTINI
>> force field. The intermolecular interactions between such molecules are
>> modelled using the Lennard-Jones potential only."
>>
>> My system is all-atom and all-hydrogen and I need to use the GBSA
>> solvation model. Is it still going to work, or is the tutorial just an
>> example of what you can do with GromPy?
>>
>> Thanks,
>>
>>      Gianluca
>>
>> On Tue, 24 Jan 2012, René Pool wrote:
>>
>>> By the way, the tutorial can be found on the wiki page at
>>> https://github.com/GromPy/GromPy/wiki/Tutorial:-Using-GromPy-in-GCMC-mode.
>>>
>>> =====================================================
>>> René Pool
>>>
>>> Division of Molecular and Computational Toxicology
>>> Department of Chemistry and Pharmaceutical Sciences
>>> Vrije Universiteit Amsterdam
>>> De Boelelaan 1083
>>> 1081HV AMSTERDAM, the Netherlands
>>> -----
>>> IBIVU/Bioinformatics
>>> Department of Computer Science
>>> Vrije Universiteit Amsterdam
>>> De Boelelaan 1081a
>>> 1081HV AMSTERDAM, the Netherlands
>>>
>>> Room P 2.75
>>> E: r.pool at vu.nl
>>> T: +31 20 598 76 12
>>> F: +31 20 598 76 10
>>> =====================================================
>>>
>>> On 01/24/2012 09:30 AM, Pool, R. wrote:
>>>       Hi Gianluca,
>>>
>>>       You might want to take a look at GromPy
>>>       (https://github.com/GromPy/GromPy).
>>>       This is a python interface to the GROMACS library. Amongst other
>>>       applications, there is an option to perform grand-canonical Monte Carlo
>>>       using GromPy. The necessary energy evaluations are performed using direct
>>>       library calls to GROMACS. In this way it is possible to get around the
>>>       possibly prohibitive file I/O in shell implementations that use the
>>>       GROMACS executables. A paper on GromPy and GromPy/GCMC has recently been
>>>       accepted in JCC, so more details will follow. In the meantime you can
>>>       have a look at the code and the tutorial on the above website.
>>>
>>>       Good luck!
>>>
>>>       Cheers,
>>>       Rene
>>>       =====================================================
>>>       René Pool
>>>
>>>       Division of Molecular and Computational Toxicology
>>>       Department of Chemistry and Pharmaceutical Sciences
>>>       Vrije Universiteit Amsterdam
>>>       De Boelelaan 1083
>>>       1081HV AMSTERDAM, the Netherlands
>>>       -----
>>>       IBIVU/Bioinformatics
>>>       Department of Computer Science
>>>       Vrije Universiteit Amsterdam
>>>       De Boelelaan 1081a
>>>       1081HV AMSTERDAM, the Netherlands
>>>
>>>       Room P 2.75
>>>       E: r.pool at vu.nl
>>>       T: +31 20 598 76 12
>>>       F: +31 20 598 76 10
>>>       =====================================================
>>>
>>> _____________________________________________________________________________________
>>> From: gmx-users-bounces at gromacs.org [gmx-users-bounces at gromacs.org] on behalf
>>> of Rodrigo Faccioli [rodrigo_faccioli at uol.com.br]
>>> Sent: 24 January 2012 05:29
>>> To: Discussion list for GROMACS users
>>> Subject: Re: [gmx-users] Monte Carlo with Gromacs and implicit solvent
>>>
>>> Hi Gianluca,
>>>
>>> I forgot to say that the step of compute the energies is considered from
>>> pdb2gmx until g_energy.
>>>
>>> Sorry about my oblivion.
>>>
>>> Best regards,
>>>
>>> --
>>> Rodrigo Antonio Faccioli
>>> Ph.D Student in Electrical Engineering
>>> University of Sao Paulo - USP
>>> Engineering School of Sao Carlos - EESC
>>> Department of Electrical Engineering - SEL
>>> Intelligent System in Structure Bioinformatics
>>> http://laips.sel.eesc.usp.br
>>> Phone: 55 (16) 3373-9366 Ext 229
>>> Curriculum Lattes - http://lattes.cnpq.br/1025157978990218
>>> Public Profile - http://br.linkedin.com/pub/rodrigo-faccioli/7/589/a5
>>>
>>>
>>> On Tue, Jan 24, 2012 at 1:07 AM, Rodrigo Faccioli <rodrigo_faccioli at uol.com.br>
>>> wrote:
>>>       Hi Gianluca,
>>>
>>>       Thanks your question.
>>>
>>>       All steps to obtain a generation with 200 individuals, on average,
>>>       is 6 minutes for 1VII, example. In each generation I must
>>>       converted my proteins Dihedral to Cartesian representation, compute
>>>       the energies, choose the individuals which will go to reproduction,
>>>       save previous population. This test was using one core of a
>>>       conventional Desktop computer.
>>>
>>>       The proteins, that I have tested my algorithm, vary from 20 to 56
>>>       amino acids.
>>>
>>>       Best regards,
>>>
>>>       --
>>>       Rodrigo Antonio Faccioli
>>>       Ph.D Student in Electrical Engineering
>>>       University of Sao Paulo - USP
>>>       Engineering School of Sao Carlos - EESC
>>>       Department of Electrical Engineering - SEL
>>>       Intelligent System in Structure Bioinformatics
>>>       http://laips.sel.eesc.usp.br
>>>       Phone: 55 (16) 3373-9366 Ext 229
>>>       Curriculum Lattes - http://lattes.cnpq.br/1025157978990218
>>>       Public Profile -
>>>       http://br.linkedin.com/pub/rodrigo-faccioli/7/589/a5
>>>
>>>
>>> On Tue, Jan 24, 2012 at 12:36 AM, Gianluca Interlandi
>>> <gianluca at u.washington.edu> wrote:
>>>       Thanks Enrico!
>>>
>>>       Just wondering. How fast is it? Calling gromacs must have a
>>>       lot of overhead. Also, do you call mdrun or does g_energy
>>>       evaluate the energy?
>>>
>>>       Gianluca
>>>
>>>       On Tue, 24 Jan 2012, Rodrigo Faccioli wrote:
>>>
>>>             Hi,
>>>
>>>             Although I don't work with MC simulation, I have
>>>             used Gromacs to obtain energies and others
>>>             protein features in my Evolutionary Algorithms
>>>             (EA).
>>>
>>>             In general lines, I created a script that calls
>>>             Gromacs programs and the output of these
>>>             programs, such as g_energy (I read its xvg file),
>>>             I stored it at a specific file which is
>>>             read by my algorithm and its values are put into
>>>             my data structure. I call Gromacs either
>>>             implicit or explicit solvent.
>>>
>>>             My project is working with EA until now. However,
>>>             the integration with GROMACS is able to
>>>             work with other kind of algorithms such as MC.
>>>             Actually, we have idea to work with MC. But,
>>>             I don't know when it will be possible. If you
>>>             want, we can talk about join these work.
>>>
>>>             Best regards,
>>>
>>>             --
>>>             Rodrigo Antonio Faccioli
>>>             Ph.D Student in Electrical Engineering
>>>             University of Sao Paulo - USP
>>>             Engineering School of Sao Carlos - EESC
>>>             Department of Electrical Engineering - SEL
>>>             Intelligent System in Structure Bioinformatics
>>>             http://laips.sel.eesc.usp.br
>>>             Phone: 55 (16) 3373-9366 Ext 229
>>>             Curriculum Lattes -
>>>             http://lattes.cnpq.br/1025157978990218
>>>             Public Profile -
>>>             http://br.linkedin.com/pub/rodrigo-faccioli/7/589/a5
>>>
>>>
>>>             On Mon, Jan 23, 2012 at 11:39 PM, Gianluca
>>>             Interlandi <gianluca at u.washington.edu> wrote:
>>>                  Hi!
>>>
>>>                  I would like to use gromacs to perform Monte
>>>             Carlo simulations in implicit
>>>                  solvent of a protein near a surface. The
>>>             protein is treated as a rigid body
>>>                  whereas the surface is fix.
>>>
>>>                  I see that there are plans to code MC into
>>>             gromacs:
>>>
>>>
>>>              http://www.gromacs.org/Developer_Zone/Programming_Guide/Monte_Carlo
>>>
>>>                  Is there a preliminary version?
>>>
>>>                  If not, I wonder whether anybody has tried
>>>             to do MC with gromacs using IMD and
>>>                  MDDriver:
>>>
>>>                  http://www.baaden.ibpc.fr/projects/mddriver/
>>>
>>>                  Besides this. Is there an easy way to obtain
>>>             the force field energy of a system
>>>                  using gromacs? Would I have to run a 0-steps
>>>             MD and read out the energy? I know
>>>                  that this would have a big overhead in a MC
>>>             simulation, but it might be worth
>>>                  trying.
>>>
>>>                  Thanks,
>>>
>>>                      Gianluca
>>>
>>>
>>>              -----------------------------------------------------
>>>                  Gianluca Interlandi, PhD
>>>             gianluca at u.washington.edu
>>>                                     +1 (206) 685 4435
>>>
>>>              http://artemide.bioeng.washington.edu/
>>>
>>>                  Postdoc at the Department of Bioengineering
>>>                  at the University of Washington, Seattle WA
>>>             U.S.A.
>>>
>>>              -----------------------------------------------------
>>>                  --
>>>                  gmx-users mailing list
>>>              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.
>>>                  Can't post? Read
>>>             http://www.gromacs.org/Support/Mailing_Lists
>>>
>>>
>>>
>>>
>>>
>>>       -----------------------------------------------------
>>>       Gianluca Interlandi, PhD gianluca at u.washington.edu
>>>                          +1 (206) 685 4435
>>>                          http://artemide.bioeng.washington.edu/
>>>
>>>       Postdoc at the Department of Bioengineering
>>>       at the University of Washington, Seattle WA U.S.A.
>>>       -----------------------------------------------------
>>>
>>> --
>>> gmx-users mailing list    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.
>>> Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>>>
>>>
>>>
>>>
>>>
>>
>> -----------------------------------------------------
>> Gianluca Interlandi, PhD gianluca at u.washington.edu
>>                     +1 (206) 685 4435
>>                     http://artemide.bioeng.washington.edu/
>>
>> Postdoc at the Department of Bioengineering
>> at the University of Washington, Seattle WA U.S.A.
>> -----------------------------------------------------
>
> -----------------------------------------------------
> Gianluca Interlandi, PhD gianluca at u.washington.edu
>                     +1 (206) 685 4435
>                     http://artemide.bioeng.washington.edu/
>
> Postdoc at the Department of Bioengineering
> at the University of Washington, Seattle WA U.S.A.
> -----------------------------------------------------

-----------------------------------------------------
Gianluca Interlandi, PhD gianluca at u.washington.edu
                     +1 (206) 685 4435
                     http://artemide.bioeng.washington.edu/

Postdoc at the Department of Bioengineering
at the University of Washington, Seattle WA U.S.A.
-----------------------------------------------------



More information about the gromacs.org_gmx-users mailing list