Re: [AMBER] The protein unfolded through the simulation

From: Thomas Cheatham <>
Date: Wed, 7 Jan 2015 21:13:34 -0700 (MST)

> I wanted to run MD simulation for a protein in a TIP3P water box with
> average 7100 atoms for 30 ns with step length of 0.001 ps. I am using NPT
> ensemble with Langevin dynamics and periodic boundary conditions.

Wow, that is an impressive unfolding event and one that likely occurs
faster than is even possible in nature... Suggests some imbalance in the
force (Luke).

An input file alone is difficult to fully interpret without context. How
was the coordinate/topology built? (e.g. what force field, what initial
coordinates?). How was the system equilibrated prior to straight /
production MD? Did you do initial minimization and then some initial MD
to get the pressure/density to relax?

Let's look at the mdin file...

> I have the following control file !
> imin=0, nmropt=0,
> ntx=5, irest=1, ntrx=1,

Restart from what? Did the initial temperature reported in the mdout look

> ntxo=1, ntpr=100, ntwr=100,
> iwrap=1, ntwx=100, ntwv=0, ntwe=0,
> ioutfm=0, ntwprt=0, idecomp=0,

We recommend using NetCDF for trajectories for better precision, ioutfm=1,
however this is not the source of the problem.

> ntf=2, ntb=2, igb=0, nsnb=1,

Constant pressure, periodic, but updating the pairlist every step?
nsnb=1 ? Not recommended, since costly, but fortunately, this is ignored
since ntb>1

In general, only specify a parameter in the mdin file if you really want
to set or change it; otherwise, let the default values help you... In
other words, only chose variables to specify that you really want to set
to values you control.

> ipol=0, gbsa=0, iesp=0,
> dielc=1.0, cut=10.0, intdiel=1.0,

cut=10.0 is overkill, but not a problem. Makes the calculation a little
slower. Could set cut=9 or 8

> ibelly=0, ntr=0,
> nstlim=5000, nscm=500, nrespa=1,
> t=0.0, dt=0.001, vlimit=20.0,

Why 1 fs time step instead of 2 fs? Did it previously blow up with a 2 fs
time step? TIP3P is fine with 2 fs under normal conditions.

> ntc=2, jfastw=0, tol=0.00001,
> Tempi=310.0 , Temp0=310.0 , ntt=3, tautp=0.5, gamma_ln=2,
> pres0=1.0, ntp=1, taup=2,

Seems OK, however why 310K? Initial velocity is ignored since you are
doing a restart. What was the initial temperature when you restarted
(printed at step 100)?

> However through the data analysis of the 30, I found that the protein
> underwent complete unfolding leaving the water box although the boundary
> conditions. Kindly find the attached snapshofts of the protein before and

It cannot leave the boundary conditions, it can span across multiple boxes
which if you visualized periodic images, you would see...

> after complete unfolding. what do you think the source of error in my
> simulation or control file ? I can support all the information that might

As is often said on this list, look at the beginning of the simulation.
Maybe even do a short simulation dumping information every step and see if
the initial energies, temperatures, densities, etc are "reasonable".

I have run a lot of simulations and never seen a protein or nucleic acid
in solution "unfold" this fast; it is likely not the mdin, unless I am
missing something but either the force field or coordinates/velocities.


AMBER mailing list
Received on Wed Jan 07 2015 - 20:30:02 PST
Custom Search