[gmx-users] Water molecule can not be settled - mdrun error

[Justin Lemkul]

On 9/24/12 4:57 PM, Lara Bunte wrote:
> Hi
>
> How to make the energy minimization in double precision? Do this mean to use grompp twice?
>

No, you would have to recompile Gromacs in a different manner:

http://www.gromacs.org/Documentation/Terminology/Precision

I don't know if we have to go that far yet.  Your forces look very bad, such 
that I don't know that a change in precision would make a difference.  Before 
taking the time to recompile Gromacs, there are other things to look at with the 
current setup.

> I changed in my energy minimization I changed emtol from 1000 to 100 and got:
>
> Steepest Descents:
>     Tolerance (Fmax)   =  1.00000e+02
>     Number of steps    =         5000
>
> Stepsize too small, or no change in energy.
> Converged to machine precision,
> but not to the requested precision Fmax < 100
>
> Steepest Descents converged to machine precision in 355 steps,
> but did not reach the requested Fmax < 100.
> Potential Energy  = -1.3305919e+05
> Maximum force     =  3.6814004e+03 on atom 19
> Norm of force     =  6.2041531e+01
>
> Is this good or bad?
>

Bad.  You're asking mdrun to converge to an Fmax < 100 and it reaches Fmax = 
3681.4 instead.  Most normal simulations need an Fmax < 1000 or so to be 
feasible, better if you can converge to a lower force.

> I also changed my .mdp file for the equilibration of the water around the flavin with the right cut-offs. Now it looks like this:
>
> define          = -DPOSRES
>
> integrator      = md
> dt              = 0.002
> emtol           = 1000.0
> emstep          = 0.01

Note that with the md integrator, you are not doing energy minimization, so 
emtol and emstep are ignored.  Not a problem per se, just pointing out 
potentially confusing settings.

> nsteps          = 5000
>
> nstlist         = 1

You need not set nstlist = 1 for running MD; this is only necessary during 
minimization.  A value of 5 or 10 would be more appropriate.

> rlist           = 1.2
> rlistlong       = 1.4
> rcoulomb        = 1.2
> coulombtype     = pme
> vdw-type        = switch
> rvdw_switch     = 0.8
> rvdw            = 1.2
>
> nstxout         = 100
> nstvout         = 100
> nstenergy       = 100
> nstlog          = 100
>
> tcoupl          = V-rescale
> tc-grps         = ISO SOL
> tau_t           = 0.1   0.1
> ref_t           = 300   300
>
> pcoupl          = no
>
> Is this okay now? I have a further question about that. As far as I understand I want to equilibrate the water around the lumiflavin which is again a kind of minimization. Why do I have to use the integrator md and not steep like in the minimization before? I use integrator md and also DPOSRES because my tutorial says I have to use this by I asked myself why not steep?
>

Energy minimization and molecular dynamics are different processes.  The "steep" 
(steepest descents) is a minimization (non-dynamical) algorithm.  It is not MD.

> Sadly with this settings I am again at the problem from the beginning of our discussion.
>
> step 266: Water molecule starting at atom 4416 can not be settled.
> Check for bad contacts and/or reduce the timestep if appropriate.
> Wrote pdb files with previous and current coordinates
>

This is a result of inadequate energy minimization.  Your starting forces are 
too high and lead to instability.  There is no point in running any MD if your 
EM does not converge adequately.

> I hope that understanding more about this minimization stuff will help me to fix this. I am interested what could be now a problem in the .mdp file.
>

The .mdp file looks reasonable (the nstlist setting should not be the cause of 
the problem), indicating to me that your topology is unsound.  I've asked about 
the source of the parameters several times, but maybe I should ask it again - 
from where did you get the parameters?  Did you derive them yourself?  Does the 
simulation run in vacuo (i.e. no water, infinite cutoffs, and no PBC)?

-Justin

-- 
========================================

Justin A. Lemkul, Ph.D.
Research Scientist
Department of Biochemistry
Virginia Tech
Blacksburg, VA
jalemkul[at]vt.edu | (540) 231-9080
http://www.bevanlab.biochem.vt.edu/Pages/Personal/justin

========================================