[gmx-users] Entropy from NMA

J. Rui Rodrigues joaquim.rodrigues at estg.ipleiria.pt
Wed Jul 28 15:03:48 CEST 2010


Dear Ran and Nahren,

I run into a similar problem a couple of weeks ago:
http://lists.gromacs.org/pipermail/gmx-users/2010-July/052377.html

In short,
"integrator =  steep " can be run in parallel. Output from parallel and 
single-node are equal.

"integrator =  l-bfgs" runs only in single-node, as advertised in the 
manual. mdrun dies with a fatal error if one tries to run in parallel:
Fatal error: Cannot do parallel L-BFGS Minimization - yet.

"integrator = nm" runs in parallel without any error. However, output 
is wrong: mdrun write multiple hessian.mtx files, each with small size.
mdrun seems to get in to some kind of race condition, because sometimes
only one  .mtx is produced (still, with small size). Running g_nmeig in these
"parallel"  mtx results in huge all-negative eigenvalues.
The same input in single-node goes just fine, with first 6 eigenvalues close
to zero.

 
Rui Rodrigues




On Wed, 28 Jul 2010 12:55:44 +0200, Ran Friedman wrote
> Your NMA seem to run on one processor four times. I'm not sure if the EM
> part can be run in parallel. This doesn't seem to be the reason for the
> negative eigenvalue. What was the g_anaeig command (with flags)?
> 
> nahren manuel wrote:
> >
> > Dear Ran,
> >
> > I ran the whole NMA again and I found where I go wrong. I hope it will
> > help others as well.
> > (sorry for a big email)
> >
> >
> > In steps 1 & 2 (of minimization) I kept the C-alpha frozen.
> >
> > 1. doublegrompp -f em1.mdp -c r1dodecapdb.pdb -p r1top.top -o em1tpr.tpr
> > mpirun -np 8 ~/gmxdpr/bin/doublemdrun -v -deffnm em1tpr
> >
> > Steepest Descents converged to Fmax < 100 in 2569 steps
> > Potential Energy  = -6.21296128286120e+03
> > Maximum force     =  9.87358378592302e+01 on atom 701
> > Norm of force     =  1.04875323563923e+01
> >
> > 2. doublegrompp -f em2.mdp -c em1tpr.gro -p r1top.top -o em2tpr.tpr -t
> > em1tpr.trr
> > mpirun -np 8 ~/gmxdpr/bin/doublemdrun -v -deffnm em2tpr
> >
> > Polak-Ribiere Conjugate Gradients converged to Fmax < 1 in 6596 steps
> > Potential Energy  = -7.56682084795001e+03
> > Maximum force     =  9.92719787231560e-01 on atom 1320
> > Norm of force     =  6.47772253797808e-02
> >
> > 3. doublegrompp -f em3.mdp -c em2tpr.gro -p r1top.top -o em3tpr.tpr -t
> > em2tpr.trr
> > mpirun -np 8 ~/gmxdpr/bin/doublemdrun -v -deffnm em3tpr
> >
> > Stepsize too small, or no change in energy.
> > Converged to machine precision,
> > but not to the requested precision Fmax < 10
> >
> > writing lowest energy coordinates.
> >
> > Steepest Descents converged to machine precision in 3018 steps,
> > but did not reach the requested Fmax < 10.
> > Potential Energy  = -8.49398064549505e+03
> > Maximum force     =  6.86336687926682e+01 on atom 935
> > Norm of force     =  7.15798864024085e+00
> >
> >
> > 4. doublegrompp -f em4.mdp -c em3tpr.gro -p r1top.top -o em4tpr.tpr -t
> > em3tpr.trr
> > mpirun -np 8 ~/gmxdpr/bin/doublemdrun -v -deffnm em4tpr
> >
> > Stepsize too small, or no change in energy.
> > Converged to machine precision,
> > but not to the requested precision Fmax < 0.01
> >
> > writing lowest energy coordinates.
> >
> >
> > Polak-Ribiere Conjugate Gradients converged to machine precision in
> > 21674 steps,
> > but did not reach the requested Fmax < 0.01.
> > Potential Energy  = -9.66058047638671e+03
> > Maximum force     =  1.49342602818840e+00 on atom 590
> > Norm of force     =  1.86576579430541e-01
> >
> >
> > 5.doublegrompp -f em5.mdp -c em4tpr.gro -p r1top.top -o em5tpr.tpr -t
> > em4tpr.trr
> >  ~/gmxdpr/bin/doublemdrun -v -deffnm em5tpr
> >
> > Low-Memory BFGS Minimizer converged to machine precision in 10346 steps,
> > but did not reach the requested Fmax < 0.0001.
> > Potential Energy  = -9.66337416273684e+03
> > Maximum force     =  6.48625019025464e-04 on atom 581
> > Norm of force     =  6.35632681200205e-05
> >
> > 6.doublegrompp -f em6.mdp -c em5tpr.gro -p r1top.top -o em6tpr.tpr -t
> > em5tpr.trr
> >
> > Low-Memory BFGS Minimizer converged to machine precision in 3316 steps,
> > but did not reach the requested Fmax < 1e-08.
> > Potential Energy  = -9.66337416275038e+03
> > Maximum force     =  6.48788529801368e-05 on atom 8
> > Norm of force     =  6.56362137550289e-06
> >
> > 7. doublegrompp -f em7.mdp -c em6tpr.gro -p r1top.top -o em7tpr.tpr -t
> > em6tpr.trr
> >
> > Low-Memory BFGS Minimizer converged to machine precision in 7530 steps,
> > but did not reach the requested Fmax < 1e-08.
> > Potential Energy  = -9.66337416275071e+03
> > Maximum force     =  4.07762128402886e-07 on atom 10
> > Norm of force     =  5.79534451487287e-08
> >
> > 8. doublegrompp -f em8.mdp -c em7tpr.gro -p r1top.top -o em8tpr.tpr -t
> > em7tpr.trr
> > Low-Memory BFGS Minimizer converged to Fmax < 1e-08 in 5840 steps
> > Potential Energy  = -9.66337416275070e+03
> > Maximum force     =  8.41750710592449e-09 on atom 521
> > Norm of force     =  1.28889193955553e-09
> >
> >
> > NMA
> >
> > 9. doublegrompp -f nma.mdp -t em8tpr.trr -c em8tpr.gro -p r1top.top -o
> > nmatpr.tpr
> > ~/gmxdpr/bin/doublemdrun -v -deffnm nmatpr
> >
> >
> > (a sinlge processor)
> >  ~/gmxdpr/bin/doublemdrun -v -deffnm nmatpr
> > it created only one nmatpr.mtx
> > Maximum force: 8.41751e-09
> > Finished step 1400 out of 1400
> >
> > It creates the output nmatpr.mtx nmatpr.edr etc....
> >
> > Now I delete all nma files
> > rm nmatpr.*
> >
> >
> > (4 processor)
> > mpirun -np 4 ~/gmxdpr/bin/doublemdrun -v -deffnm nmatpr
> > Maximum force: 8.41768e-09
> > Maximum force: 8.41768e-09
> > Maximum force: 8.41768e-09
> > Maximum force: 8.41768e-09
> > Finished step 1399 out of 1400
> >
> > Writing Hessian...
> >
> >
> > Writing Hessian...
> >
> >
> > Writing Hessian...
> > Finished step 1400 out of 1400
> >
> > Writing Hessian...
> >
> > Back Off! I just backed up nmatpr.mtx to ./#nmatpr.mtx.2#
> >
> > It gives me negative eigenvalues.
> >
> > One would expect it should not create nmatpr.mtx.1 nmatpr.mtc.2
> > etc...since these files have been removed. Because it created multiple
> > nmatpr.mtx file and backup one over another
> >
> > I am not sure If the problem occurs with my system , but other MPIRUN
> > (with mdrun) do not create such backup.
> >
> > Best,
> > nahren
> >
> >
> > --- On *Wed, 7/28/10, Ran Friedman /<r.friedman at bioc.uzh.ch>/* wrote:
> >
> >
> >     From: Ran Friedman <r.friedman at bioc.uzh.ch>
> >     Subject: Re: [gmx-users] Entropy from NMA
> >     To: "Discussion list for GROMACS users" <gmx-users at gromacs.org>
> >     Date: Wednesday, July 28, 2010, 3:10 PM
> >
> >     p.s. How big is this negative value?
> >
> >     Ran Friedman wrote:
> >>     Hi,
> >>
> >>     Can you post the exact commands you used for EM and NMA?
> >>
> >>     Ran
> >>
> >>     nahren manuel wrote:
> >>>     Dear Gromacs Users,
> >>>
> >>>     I am trying to calculate Entropy from Normal Mode Analysis.
> >>>     I minimized the structure to
> >>>     8.41750710592449e-09
> >>>
> >>>     Then I calculated the normal mode (the mdp file is given below).
> >>>
> >>>     But when i try to calculate the Eigenvalues, I get the following
> >>>     warning
> >>>
> >>>
> >>>     One of the lowest 6 eigenvalues has a non-zero value.
> >>>     This could mean that the reference structure was not
> >>>     properly energy minimized.
> >>>     Writing eigenvalues...
> >>>
> >>>     I get only negative eigenvalues.
> >>>
> >>>     Let me also add that I used the same mdp file for minimization
> >>>     changing the integrator to steep, cg and finally to l-bfgs.
> >>>
> >>>     Best,
> >>>     nahren
> >>>
> >>>     -----------------------
> >>>     integrator    = nm     
> >>>     constraints = none
> >>>     define = -DFLEXIBLE
> >>>     emtol        = 0.00000001    
> >>>     emstep          = 0.00000001     
> >>>     nsteps        = 50000       
> >>>     nstlist        = 1      
> >>>     ns_type        = grid    
> >>>     rlist        = 1.5     
> >>>     coulombtype    = shift     
> >>>     rcoulomb-switch    = 1.0    
> >>>     rvdw-switch        = 1.0      
> >>>     vdwtype = shift
> >>>     rcoulomb = 1.3
> >>>     rvdw = 1.3
> >>>
> >>>
> >>
> >>
> >>     -- 
> >>     ------------------------------------------------------
> >>     Ran Friedman
> >>     Postdoctoral Fellow
> >>     Computational Structural Biology Group (A. Caflisch)
> >>     Department of Biochemistry
> >>     University of Zurich
> >>     Winterthurerstrasse 190
> >>     CH-8057 Zurich, Switzerland
> >>     Tel. +41-44-6355559
> >>     Email: r.friedman at bioc.uzh.ch </mc/compose?to=r.friedman at bioc.uzh.ch>
> >>     Skype: ran.friedman
> >>     ------------------------------------------------------
> >>       
> >
> >
> >
> >     -----Inline Attachment Follows-----
> >
> >     -- 
> >     gmx-users mailing list    gmx-users at gromacs.org
> >     </mc/compose?to=gmx-users at gromacs.org>
> >     http://lists.gromacs.org/mailman/listinfo/gmx-users
> >     Please search the archive at http://www.gromacs.org/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
> >     </mc/compose?to=gmx-users-request at gromacs.org>.
> >     Can't post? Read http://www.gromacs.org/mailing_lists/users.php
> >
> >
> 
> -- 
> ------------------------------------------------------
> Ran Friedman
> Postdoctoral Fellow
> Computational Structural Biology Group (A. Caflisch)
> Department of Biochemistry
> University of Zurich
> Winterthurerstrasse 190
> CH-8057 Zurich, Switzerland
> Tel. +41-44-6355559
> Email: r.friedman at bioc.uzh.ch
> Skype: ran.friedman
> ------------------------------------------------------


--
Webmail ESTG de Leiria (http://webmail.estg.ipleiria.pt)




More information about the gromacs.org_gmx-users mailing list