[AMBER] Problems with MC barostat

From: Yessica <yessica.gomez.ucsf.edu>
Date: Sun, 12 Jul 2020 16:53:08 -0700

Dear AMBER folks,

I have been doing fully atomistic simulations of a protein in a POPC
bilayer using AMBER18 with the AMBER forcefields. The system has
~305,000 atoms and box dimensions of 200x200x80Å.
I have repeatedly encountered a problem where after ~100ns of
unrestrained (production) simulation it becomes apparent that the box is
shrinking in the X and Y dimensions, while growing in the Z dimension.
This behavior is seen for multiple systems of the same size. For
production, I use the following parameters:

A NPT simulation for common production-level simulations
     imin=0, ! No minimization
     irest=1, ! This IS a restart of an old MD simulation
     ntx=5, ! So our inpcrd file has velocities

     ! Temperature control
     ntt=3, ! Langevin dynamics
     gamma_ln=1.0, ! Friction coefficient (ps^-1)
     temp0=303.15, ! Target temperature

     ! Potential energy control
     cut=10.0, ! Nonbonded cutoff, in Angstroms

     ! MD settings
     nstlim=100000000, ! 100M steps, 200 ns total
     dt=0.002, ! time step (ps)

     ! SHAKE
     ntc=2, ! Constrain bonds containing hydrogen
     ntf=2, ! Do not calculate forces of bonds containing hydrogen
     tol=0.0000001, ! SHAKE tolerance

     ! Control how often information is printed
     ntpr=100000, ! Print energies every 100K steps
     ntwx=100000, ! Print coordinates every 100K steps to the trajectory
     ntwr=100000, ! Print a restart file every 100K steps (can be less frequent)
     ntxo=2, ! Write NetCDF format
     ioutfm=1, ! Write NetCDF format (always do this!)

     ! Wrap coordinates when printing them to the same unit cell

     ! Constant pressure control.
     barostat=2, ! MC barostat... change to 1 for Berendsen
     ntp=3, ! 1=isotropic, 2=anisotropic, 3=semi-isotropic w/ surften
     pres0=1.0, ! Target external pressure, in bar
     taup=1.0, ! Was 10 in Yessica's mineq but has been reset to default here

     ! Constant surface tension (needed for semi-isotropic scaling). Uncomment
     ! for this feature. csurften must be nonzero if ntp=3 above
     csurften=3, ! Interfaces in 1=yz plane, 2=xz plane, 3=xy plane
     gamma_ten=0.0, ! Surface tension (dyne/cm). 0 gives pure semi-iso scaling
     ninterface=2, ! Number of interfaces (2 for bilayer)
     ntb=2, ! pbc
   &wt type='END' / ! End of varying conditions

However, if I start a parallel simulation with the same parameter files
using the Berendsen barostat instead of MC (only change is barostat=1),
I do NOT observe the box compression even at 200+ ns of run-time. I
believe this also speaks to the question raised by David Case in a
recent thread: http://archive.ambermd.org/202006/0010.html

Why might this be happening, and is there a way I can fix it?

AMBER mailing list
Received on Sun Jul 12 2020 - 17:00:03 PDT
Custom Search