Hi,
I tried to output the gradient and hessian from nmode.c. Again, for an isolated molecule the gradients from sander and NAB agree well very (diff < 0.1% or less ). Howvever, for the molecule in solvent, the force from NAB is order of magnitude larger than those from sander (microcanonical MD with PBC, ntb=1). My input for NAB and sander are listed below (I set dgrad to a large number to avoid minimization). I also tried to add option bcopt=10 (to use PBC) but the computed force from NAB showed no changes. Did I make some stupid mistakes here? Thanks!
Best,
JT
------------- Input for NAB ---------------------
// Try some molecular mechanics.
molecule m;
int ier;
float m_xyz[dynamic], f_xyz[dynamic];
float dgrad, fret;
m = getpdb("Solvated.pdb");
allocate m_xyz[3*m.natoms];
allocate f_xyz[3*m.natoms];
readparm(m, "Solvated.prmtop");
mm_options("cut=10.0, diel=C");
mme_init(m, NULL, "::ZZZZ", m_xyz, NULL);
// Do some conjugate gradients.
dgrad = 50000000000.055;
setxyz_from_mol( m, NULL, m_xyz );
ier = conjgrad(m_xyz, 3*m.natoms, fret, mme, dgrad, 1.0, 1);
// Do some Newton-Raphson.
mm_options("ntpr=1,cut=10.0, diel=C");
dgrad = 1.e+16;
ier = newton(m_xyz, 3*m.natoms, fret, mme,mme2, dgrad, 0.0, 1);
// get the normal modes:
ier = nmode( m_xyz, 3*m.natoms,mme2, 0,1,0.0,0.0,0);
if(mytaskid==0)
printf("nmode returns %d\n", ier );
------- Input for sander -----------
&cntrl
imin = 0, ! Run molecular dynamics without any minimization
irest = 0, !
ntx = 1, ! read coordinate, but no velocities
ntf = 1, ! complete interacion evaluation
ntb = 1, ! constant volume
dielc = 1.0, ! Dielectric multiplicative constant for the electrostatic interactions
cut = 10.0, ! the nonbonded cutoff, in Angstroms
ntc = 1, ! No SHAKE
ntt = 0, ! Constant total energy classical dynamics
nstlim = 1, ! number of MD step
dt = 0.0001,
ntpr = 1,
ntwx = 1,
ntwr = 10000,
/
________________________________
From: David A Case <case.biomaps.rutgers.edu>
To: John Travers <jtravers70.yahoo.com>; AMBER Mailing List <amber.ambermd.org>
Sent: Saturday, August 10, 2013 9:30 AM
Subject: Re: [AMBER] output nuclear force and hessian
On Fri, Aug 09, 2013, John Travers wrote:
>
> I had another strange observation for the forces computed by nmode
> and sander. For a molecule in solvent, I performed a microcanonical
> dynamics. I found the the forces on the solute molecule obtained from
> these two code agree reasonably well, the difference < 10%. However,
> the forces on solvent molecules show large disparity, often > 50%. Are
> these difference still resulted from the different implementation and
> different algorithms? Thanks!
You don't give many details about what you did, but here is a guess: If the
"molecule in a solvent" calculation used periodic boundary conditions, then
what you see is to be expected: nmode has no capability to handle periodicity,
and will give incorrect results if the simulation was run with PBC turned on.
...dac
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Sat Aug 10 2013 - 16:00:03 PDT
