[gmx-users] Re: SMD - reproducibility

[André Farias de Moura]

there should be an extra option in your mdrun command line if
reproducibility is an issue:

-[no]reprod  bool   no      Try to avoid optimizations that affect binary
reproducibility

the value of -reprod is set to no by default, turn on that option and
trajectories should be the same.

best

Andre

On Fri, Apr 26, 2013 at 9:25 AM, Steven Neumann <s.neumann08 at gmail.com>wrote:

> Thanks anyway. I will try tu pull it 20-40 times to tell something.
>
> Steven
>
> On Fri, Apr 26, 2013 at 1:23 PM, Thomas Schlesier <schlesi at uni-mainz.de
> >wrote:
>
> > Don't know. One idea i have: Take a flexible and a relative rigid system
> > and perform simulations with the same starting conditions (-> using -t
> > *.cpt). I would imagine that for the flexible system the trajectories
> start
> > earlier to deviate, since more stuff could happen (system is more
> flexible
> > -> greater configurational space). For the rigid system the
> configurational
> > space is smaller, so the probability is higher to always follow the same
> > trajectory if one starts with a predefinded velocity and direction.
> > But don't know if this is true, but it's the first thing which comes to
> my
> > mind.
> >
> >
> > Am 26.04.2013 14:01, schrieb gmx-users-request at gromacs.org:
> >
> > Thanks for this. I think option 2 is more reasonable. However, still do
> not
> >> know why I get sometimes 3 types of profiels and sometimes 10 for 10 SMD
> >> simulations...
> >>
> >> On Fri, Apr 26, 2013 at 12:46 PM, Thomas Schlesier<schlesi at uni-mainz.de
> **
> >> >wrote:
> >>
> >> >Think i now understand your question. Forget what i wrote before.
> >>> >I could imagine the the 'grompp -t npt.cpt' part is a problem.
> >>> >If the simulations would be numerical reproducible, one should get the
> >>> >same results. As they are not, the results will differ somewhat (would
> >>> >think the more, the longer you simulate). But two different
> simulations
> >>> >would be more equal to each other, than two simulations which start
> with
> >>> >different velocity distributions for the particles.
> >>> >
> >>> >
> >>> >If you're interested in an stochastic analysis of your system (meaning
> >>> >simulations which are not equal - performing many pulling experiments
> in
> >>> >reality, one would also have many different starting points) you could
> >>> do
> >>> >two things:
> >>> >1) Run a look npt simulation, and use different frames to start the
> SMD
> >>> >simulations. From each frame the particles should have a different
> >>> velocity
> >>> >distribution and the results of the SMD simulations should also
> differ.
> >>> >(Depending on how many SMD simulations you want to perform, this might
> >>> get
> >>> >expensive, since the starting frame for SMD should be separated by
> more
> >>> >then a few ps.)
> >>> >2) Dump the 'npt.cpt' file and randomly determine new velocities for
> >>> each
> >>> >particle at the start of each SMD simulation. Since each simulation
> has
> >>> a
> >>> >different velocity distribution, the SMD simulation wont be the same.
> >>> This
> >>> >approach has only one weak point. Due to assigning new random
> velocities
> >>> >you destroy the thermal equilibrium of the system. But if the system
> was
> >>> >well equilibrated before, this distrubance should only be small and
> >>> after
> >>> >the first 100-200 ps of the SMD simulaton the system is in thermal
> >>> >equilibrium. If the complete SMD simulation is much longer (couple of
> >>> ns),
> >>> >the interesting stuff would happen longer after the inital simulation
> >>> time
> >>> >with the destroyed equilibrium.
> >>> >
> >>> >Hope this helps
> >>> >Thomas
> >>> >
> >>> >
> >>> >
> >>> >Am 26.04.2013 12:00, schriebgmx-users-request@**gromacs.org<
> schriebgmx-users-request at gromacs.org>
> >>> :
> >>> >
> >>> >Dear Users,
> >>>
> >>>> >>
> >>>> >>I am running my puling simulations of ligand with constant velocity.
> >>>> First
> >>>> >>I minimize and equilibrate my system:
> >>>> >>
> >>>> >>grompp -f minim.mdp -c Solvions.gro -p topol.top -o em.tpr
> >>>> >>mdrun -s em.tpr -deffnm em
> >>>> >>grompp -f nvt298US.mdp -n index.ndx -c em.gro -p topol.top -o
> >>>> nvt298.tpr
> >>>> >>mdrun -s nvt298.tpr -deffnm nvt298
> >>>> >>grompp -f npt298US.mdp -n index.ndx -c nvt298.gro -t nvt298.cpt -p
> >>>> >>topol.top -o npt298.tpr
> >>>> >>mdrun -s npt298.tpr -deffnm npt298
> >>>> >>
> >>>> >>Then I run 10 pulling simulations with the same mdp file:
> >>>> >>
> >>>> >>grompp -f pull.mdp -c npt298.gro -p topol.top -n index.ndx -t
> >>>> npt298.cpt
> >>>> >>-o
> >>>> >>pull_1.tpr
> >>>> >>mpiexec mdrun -s pull_1.tpr -deffnm pull_1
> >>>> >>
> >>>> >>...
> >>>> >>
> >>>> >>grompp -f pull.mdp -c npt298.gro -p topol.top -n index.ndx -t
> >>>> npt298.cpt
> >>>> >>-o
> >>>> >>pull_10.tpr
> >>>> >>mpiexec mdrun -s pull_10.tpr -deffnm pull_10
> >>>> >>
> >>>> >>
> >>>> >>I get 3 different (but similar) profiles (Force vs time) with 10
> >>>> >>simulations as some of them produce exactly the same results... In
> >>>> another
> >>>> >>system with the same methodology I get 10 similar but different
> >>>> profiles.
> >>>> >>I
> >>>> >>am wondering why in this case only 3 types are possible... Shall I
> try
> >>>> >>grompp without -t npt.cpt ?
> >>>> >>
> >>>> >>Steven
> >>>> >>
> >>>>
> >>>
> > --
> > gmx-users mailing list    gmx-users at gromacs.org
> > http://lists.gromacs.org/**mailman/listinfo/gmx-users<
> http://lists.gromacs.org/mailman/listinfo/gmx-users>
> > * Please search the archive at http://www.gromacs.org/**
> > Support/Mailing_Lists/Search<
> 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<
> http://www.gromacs.org/Support/Mailing_Lists>
> >
> --
> 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
>
>


-- 
_____________

Prof. Dr. André Farias de Moura
Department of Chemistry
Federal University of São Carlos
São Carlos - Brazil
phone: +55-16-3351-8090