Re: PMEGridSize

From: JC Gumbart (gumbart_at_ks.uiuc.edu)
Date: Thu Jul 23 2009 - 19:08:52 CDT

You will not decrease simulation time by using rigidbonds with a 1fs
timestep. You'll actually increase it, due to the cost of the shake
algorithm. It's not like the hydrogens are ignored when the bonds
are rigid. A few quick tests will show you the difference in
computational time.

On Jul 23, 2009, at 3:13 PM, Rabab Toubar wrote:

> Hi,
> When there is no need to use Rigidbonds all with a 1 fs, "can" we
> use it? Just to decrease the computer cost and the simulation time.
> Thanks
>
> Rabab Toubar
>
>
> --- On Sat, 10/11/08, JC Gumbart <gumbart_at_ks.uiuc.edu> wrote:
>
> From: JC Gumbart <gumbart_at_ks.uiuc.edu>
> Subject: RE: namd-l: PMEGridSize
> To: rtoubar_at_yahoo.com, char_at_ks.uiuc.edu
> Cc: namd-l_at_ks.uiuc.edu
> Date: Saturday, October 11, 2008, 2:01 AM
>
> Thereís no need to use shake (rigidbonds) with a 1 fs timestep.
> See the description of multiple time stepping here: http://
> www.ks.uiuc.edu/Research/namd/2.6/ug/
> node26.html#SECTION00083700000000000000
>
>
> Also, I guess the NAMD tutorial will discuss in detail as well
> along with sample parameters.
>
>
> As was alluded to by Victor, the multiple time stepping algorithm
> does not conserve energy exactly, although Iím not sure it would
> cause a sharp drop. Try it (fullElectFreq 1) though to see if this
> eliminates your problem (keeping in mind your simulation will run
> slower then). Additionally, your switch distance is too low; it
> should be 10 A, not 8. The namd manual gives a pretty good
> description of what this is and what it should be set to.
>
>
> Regarding the protein leaving the box, this could just be the
> natural diffusion, combined with how VMD wraps molecules around
> periodic boundaries. Presumably, there is a hole on the side of
> the box opposite to where the protein is?
>
>
>
> From: owner-namd-l_at_ks.uiuc.edu [mailto:owner-namd-l_at_ks.uiuc.edu] On
> Behalf Of Rabab Toubar
> Sent: Friday, October 10, 2008 8:32 AM
> To: char_at_ks.uiuc.edu
> Cc: namd-l_at_ks.uiuc.edu
> Subject: Re: namd-l: PMEGridSize
>
>
> Thanks Chris for your reply,
>
>
> Below is my configuration file for the equilibration step. It is
> worth mentioning that I tried increasing the PMEGridsize and that's
> when my equilibration went fine. Visualizing the result using VMD,
> my molecule is leaving the solvation box completely.
>
>
> Any suggestions for what could be wrong?
>
> Thanks
>
> Rabab
>
>
> --------------------
>
> # input topology and initial structure..............
> structure dfn_solv.psf
> coordinates ./output/dfn_heat.coor
>
> #..force field block ........................
> paratypecharmm on
> parameters par_all22_prot.inp
> exclude
> scaled1-4
> 1-4scaling 1.0
> dielectric 1.0
>
> # dealing with long-range
> interactions..............
> switching on
> switchdist 8.0
>
> cutoff 12.0
> pairlistdist
> 13.5
> margin 0.0
> stepspercycle
> 20
> rigidBonds all
>
> rigidTolerance 0.00001
> rigidIterations 100
>
> # this block specifies the Ewald
> electrostatics.........................
> PME
> on
>
> PMETolerance 0.000001
>
> PMEGridSizeX
> 32
> PMEGridSizeY 32
> PMEGridSizeZ 32
>
> #block specifying the parameters for integrator and MTS
> timestep 1.0
> fullElectFrequency
> 4
>
> # this block specifies the output....
> outputenergies
> 1000
>
> outputtiming 1000
>
> binaryoutput no
>
> outputname output/dfn_equil
>
> restartname output/dfni_equil
>
> restartfreq 10000
> binaryrestart yes
> DCDfile output/dfn_equil.dcd
> dcdfreq 1000
>
> #MD protocol block
> ...................
> seed 1551
> numsteps 4000000
> temperature 300
>
> rescaleFreq 1
> rescaleTemp 300
>
>
> # this block defines periodic boundary conditions......
> cellBasisVector1 37.0 0.0 0.0
> cellBasisVector2 0.0 31.0 0.0
> cellBasisVector3 0.0 0.0 30.3
>
> cellOrigin 0.0 00.0 0.0
> wrapWater on
>
>
> --- On Fri, 10/10/08, Chris Harrison <char_at_ks.uiuc.edu> wrote:
>
> From: Chris Harrison <char_at_ks.uiuc.edu>
> Subject: Re: namd-l: PMEGridSize
> To: rtoubar_at_yahoo.com
> Cc: namd-l_at_ks.uiuc.edu
> Date: Friday, October 10, 2008, 1:37 AM
>
> Rabab,
>
> Post your namd configuration file. You have an energy leak.
> Assuming your system (the initial structure) is not pathological,
> we'll need to see your simulation parameters to judge the leak's
> cause.
>
> And to clarify, *what* "went through to the end"? The
> equilibration simulation or the entire protocol of simulations?
>
> Chris
>
>
>
> On Mon, Oct 6, 2008 at 10:51 AM, Rabab Toubar <rtoubar_at_yahoo.com>
> wrote:
>
> I do the minimization, heating and then equilibration steps
> separately before the simulation step. All ran fine apart from the
> simulation step that terminated without any error messages after
> may be 1/3 the simulation time. I had 2 suggestions: 1- PMEGridsize
> is too small and need to be > CellBasisVector (and this is
> mentioned in the tutorial) - 2 - the system didn't equilibrate ..so
> I need to run the equilibration step long enough.
>
> Adjusting the PME Gridsize and rerunning the simulation, it went
> through to the end BUT plotting the Total Energy vs TS showed a
> sharp decrease when it is supposed to be constant.
>
> Any suggestions?
> Thanks
> Rabab
>
> --- On Fri, 10/3/08, JC Gumbart <gumbart_at_ks.uiuc.edu> wrote:
>
> From: JC Gumbart <gumbart_at_ks.uiuc.edu>
>
>
> Subject: Re: namd-l: PMEGridSize
> To: rtoubar_at_yahoo.com
>
> Cc: "Roman Petrenko" <rpetrenko_at_gmail.com>, tutorial-l_at_ks.uiuc.edu,
> namd-l_at_ks.uiuc.edu
> Date: Friday, October 3, 2008, 10:53 PM
>
>
> Many shortcuts are taken in the NAMD tutorial simulations in the
> interest of speed; I'm not sure how much the PME grid size affects
> speed though (you could check yourself if interested). Therefore,
> what's written in the tutorial is much more important than what's
> in the configuration files themselves. While the simulation may
> appear to run fine with a coarse PME grid for a while, it's not
> accurate.
>
>
> Regarding why your simulation terminated unexpectedly, you will
> need to give us more information to determine that.
>
>
> On Oct 3, 2008, at 3:22 PM, Rabab Toubar wrote:
>
>
>
>
> So you are suggesting leaving the Grid Size as (32s) is and doing
> this gradual heating?
>
>
> Thanks Roman,
>
> Rabab
>
> --- On Thu, 10/2/08, Roman Petrenko <rpetrenko_at_gmail.com> wrote:
>
> From: Roman Petrenko <rpetrenko_at_gmail.com>
> Subject: Re: namd-l: PMEGridSize
> To: rtoubar_at_yahoo.com
> Cc: tutorial-l_at_ks.uiuc.edu, namd-l_at_ks.uiuc.edu
> Date: Thursday, October 2, 2008, 4:22 PM
>
> gradual heating solved similar problem in my case. put this at the end
>
> of your namd.conf script:
>
> ----------------------------------------------------------------------
> -------------------
>
> set temperature 310
>
>
>
>
> set minimize_steps 300
>
> set run_steps 10000000
>
> #############################################################
>
> ## EXECUTION SCRIPT ##
>
> #############################################################
>
>
>
>
>
>
> # Minimization
>
> minimize $minimize_steps
>
> reinitvels
> $temperature
>
>
>
> for {set i 25} {$i < $temperature} {incr i 25} {
>
> set tempr $i
>
> langevinTemp $tempr
>
> reinitvels $tempr
>
> run 100
>
>
>
>
> }
>
>
>
> firsttimestep 0
>
> langevinTemp $temperature
>
> run $run_steps
>
>
>
>
>
>
>
> On Thu, Oct 2, 2008 at 2:55 PM, Rabab Toubar <rtoubar_at_yahoo.com>
> wrote:
>
>
>
>
> >
> Hi,
>
> >
>
> > In the namd tutorial it says on page 24: " Typically, a grid
>
> size
>
> > slightly less than 1 is a
> good
>
> > A
>
> > choice to reproduce charge distribution in biological systems,
>
>
>
>
> > where the closest atoms have a bond separation on the order of 1
>
> > A. Furthermore, for speed in computing Fast Fourier Transforms,
>
> > PMEGridSizeX should be chosen so that it can be factorized by
>
> > 2, 3, or 5. If your cellBasisVector1 = (60, 0, 0), a good
>
>
>
>
> > choice for PMEGridSizeX might be 64, since 60 A/ 64 = 0.9375
>
> > A and 64 = 26."
>
> >
>
> > While in the namd tutorial they used gridsizes
> of 32 in each direction
>
> when
>
> > the cellBasisVector was 42, 44 and 47, yet the simulation went OK.
>
>
>
>
> >
>
> > I am asking this as I was running a simulation of cellBasisVector
> around
>
> 35
>
> > in each direction, and the gridsizes were
> set to 32s according to the
>
> > tutorial. I have been looking for the reasons of the unexpected
>
> termination
>
> > of the simulation. And according to the tutorial this couldn't be
> one.
>
> >
>
> > * My question is, how far could the PMEGridSizes affect the
> simulation?
>
>
>
>
> > * And if the equilibration step wasn't long enough, could this
>
> terminate the
>
> > simulation before completion?
>
> >
>
> > Thanks
>
> > Rabab
>
> >
>
> >
>
>
>
>
>
>
>
> --
>
> Roman Petrenko.
>
>
>
>
>
> Physics Department
>
> University of Cincinnati
>
>
>
>
>
> --
> Chris Harrison, Ph.D.
> Theoretical and Computational Biophysics Group
> NIH Resource for Macromolecular Modeling and Bioinformatics
> Beckman Institute for Advanced Science and Technology
> University of Illinois, 405 N. Mathews Ave., Urbana, IL 61801
>
> char_at_ks.uiuc.edu Voice: 217-244-1733
> http://www.ks.uiuc.edu/~char Fax: 217-244-6078
>
>
>
>

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:53:04 CST