AMBER Archive (2004)

Subject: Re: AMBER: GB and langevin dynamics

From: Sergio E. Wong (swon9_at_itsa.ucsf.edu)
Date: Wed Sep 15 2004 - 14:38:43 CDT

I lowered the relative geometrical tolerance for coordinate resetting in
shake, set igb=0 and increased nscm to be longer than the simulation. I
learned
that:

1) using constant energy dynamics, the energy of my system is stable.
2) using the Andersen thermostat (vrand 500 and no center of mass velocity
removal), the average temperature matches the target temperature (over
several nanoseconds). Using the Berensen weak-coupling scheme, the
average temperature also matches the target temperature.
3) if I set nscm to 5000 when using the Andersen thermostat, the average
temperature is about 10 degrees lower (~290 vs 300). The low number of
degrees of freedom in this system probably make it particularly sensitive
to this change in velocities.
4) taking the input script from the successful runs reassigning velocities
(Andersen) or using the Berendsen scheme, and simply setting gamma_ln=2
and ntt=3, still causes the average temperature to be higher (310-315)
than the target temperature.

5) Running langevin dynamics on a larger system (NATOMS ~ 3700) using the
same mdin file produces the right average temperature (even over a short
25000 step simulation).

My input mdin file is:

 &cntrl
  imin = 0,
  nmropt = 0,

  ntx = 5, irest = 1,

  ntpr = 1000, ntwx = 1000,
  ntwr = 1000,

  ntf = 2, ntb = 0, cut = 50,

  igb = 0, GAMMA_LN=2,

  nstlim = 500000, t = 0.0, dt = 0.001,

  temp0 = 300.0, tempi = 300.0,
  ntt = 3, dtemp = 0.0,
  nscm= 50000000,

  ntp = 0, ntc = 2, tol = 0.0000001,
  vrand = 5000,

 &end

I produced the topology and crd files in tleap doing the following:

source leaprc.ff96
set default PBradii mbondi2
mol = sequence {ACE GLY SER SER SER NME}
saveamberparm mol pep.top pep.crd
quit

thanks

-Sergio..

On Mon, 13 Sep 2004, David A. Case wrote:

> On Thu, Sep 09, 2004, Sergio E. Wong wrote:
> >
> >
> > I'm concerned b/c I ran a GB + langevin dynamics simulation of a short
> > peptide (ACE GLY SER SER SER NME) and the average temperature (over
> > 500,000 steps = 0.5 ns) is about 10 degrees above the target temperature.
>
> I haven't seen this in simulations we have done. You might check that
> energy is conserved wtih temperature coupling turned off (ntt=0) for your
> system. Also worth checking is the value of nscm you use (try a value of
> 1000 or 2000), to see if that makes a difference. You could also try
> increasing gamma_ln, but this shouldn't make a difference, and doesn't in
> my test cases.
>
> In the end, we may have to get your input files to see what is really
> happening...you are correct to be concerned about this sort of result.
>
> ...good luck...dac
>
>
> -----------------------------------------------------------------------
> The AMBER Mail Reflector
> To post, send mail to amber_at_scripps.edu
> To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu
>
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber_at_scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu