AMBER Archive (2009)

Subject: [AMBER] nab conjgrad min problem

From: Jeremy Harris (jphillipharris_at_gmail.com)
Date: Fri Jan 16 2009 - 09:11:12 CST

Thanks for the previous help - you recommended the NMA from nab over nmode.
I've been fiddling around with nab, and can't seem to get it to do a simple
conjgrad minimization. On the first step it will give an enormously large
negative energy, and then change the position of one of the atoms by an
equally large number. Could you also explain the fourth argument in
mme_init?

my min.nab file and output are:

molecule p;
float x[ 4000 ], fret, dummy[2];

p = getpdb("run19.pdb");
readparm(p, "run19.prmtop");

mm_options("cut=999, ntpr=1, nsnb=9999, diel=R94, gb=0, temp0=300.0");
mme_init( p, NULL, "::ZZZ", dummy, NULL);
setxyz_from_mol( p, NULL, x);

//conj grad min
conjgrad(x, 3*p.natoms, fret, mme, 0.1, 10.0, 1);
//conjgrad(float x[], int n, float fret, func, float rmsgrad, dfpred,
maxiter)
setmol_from_xyz(p, NULL, x);
putpdb("cg.min.pdb",p);

output:

Reading parm file (run19.prmtop)
title:

        mm_options: cut=999
        mm_options: ntpr=1
        mm_options: nsnb=9999
        mm_options: diel=R94
        mm_options: gb=0
        mm_options: temp0=300.0
      iter Total bad vdW elect nonpolar genBorn
frms
ff: 0 -456.84 3525.32 -591.83 -3390.33 0.00 0.00
1.82e+01
ff: 1 120962626.15 120966596.09 -590.70 -3379.24 0.00 0.00
8.44e+03

pdb output, atom 1334:
ATOM 1334 O ALA 88 -1.260120962626.148 38.096 0.00 0.00

Thanks again,
Jeremy Harris

On Wed, Jan 14, 2009 at 1:13 PM, David A. Case <case_at_biomaps.rutgers.edu>wrote:

> On Tue, Jan 13, 2009, Jeremy Harris wrote:
> >
> > I'm just curious now about the time it takes for minimization. It is
> taking
> > about 15 min per minimization step, is this normal? The protein has 150
> > residues.
>
> That sounds about right for Newton Raphson. You might want to use the
> xmin minimizer to get to a low RMS gradient (below 10**-5) before
> starting NR. Also, bear in mind that nmode is very old code, and that
> NAB allows you to do normal modes with GB (rather than just vacuum), and
> in a code environment that is still being maintained.
>
> ...dac
>
>
> _______________________________________________
> AMBER mailing list
> AMBER_at_ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>
_______________________________________________
AMBER mailing list
AMBER_at_ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber