AMBER Archive (2009)

Subject: Re: [AMBER] amber8 pmemd trajectory coordinates problem

From: Robert Duke (rduke_at_email.unc.edu)
Date: Tue Oct 20 2009 - 20:15:56 CDT

Oh sorry, I missed the comment about amber 8 vs. 10. I am not aware of any
differences that would cause this; I would first suspect it is a phasing
issue between the two runs, though I have no idea how that could happen (the
respa counter starts at the beginning of each run). If I were doing the
simulations and trying to just get more throughput per cpu hour, I would
first consider using pmemd 10, which should be significantly faster than 8;
if I had to have more throughput still, I would consider using a 2 fsec step
instead of a 1 fsec step. ALSO, and this is a big ALSO, using a cut of 12
is very expensive in pme and should be avoided (the direct force time will
be about 12**3 / 8**3 = 3x more expensive, so all that effort to save time
on the reciprocal calcs was wasted). If you really want to run fast, but
want better vdw cutoffs than the default (default cut = 8 for both vdw and
direct electrostatics), I would use es_cutoff = 7 or 8, and vdw_cutoff = 9;
this gives pretty good vdw and really cheap but still effective pme
electrostatics.
Regards - Bob Duke
----- Original Message -----
From: <wong105_at_llnl.gov>
To: <amber_at_ambermd.org>
Sent: Tuesday, October 20, 2009 8:32 PM
Subject: [AMBER] amber8 pmemd trajectory coordinates problem

Dear Amber users;

     I ran an explicit water simulation (TIP3P, PBC, counter ions,
pme, etc) using PMEMD version 8. The problem is that when I visualize
the trajectory it seems to advance very little except every 4th
trajectory step. The system is of medium size ( 48732 Atoms) and my
input file is below:

  &cntrl
    imin = 0,
    ntx=5,
    irest=1,
    ntpr = 5000,
    ntwx = 5000,
    iwrap=1,
    ntwr = 5000,
    nstlim =1000000,
    dt = 0.001,
    ntb = 1,
    cut = 12,
    ntr = 0,
    ntc = 2,
    ntf = 2,
    ntt = 1,
    temp0 = 300.0,
    nrespa=4,
  &end

To illustrate what I mean, I ran a rms fit to C-alphas analysis using
ptraj. The output is below. I placed stars at positions where the
C-alpha rmsd changed significantly from the previous.

     4.00 0.82920
     5.00 0.82906
     6.00 0.82890
     7.00 0.82872
     8.00 0.82843
     9.00 0.82844
    10.00 0.82839
    11.00 0.82832
    12.00 0.86168 *****
    13.00 0.86181
    14.00 0.86193
    15.00 0.86208
    16.00 0.91274 *****
    17.00 0.91281
    18.00 0.91294
    19.00 0.91309
    20.00 1.06776 *****
    21.00 1.06770
    22.00 1.06768
    23.00 1.06790
    24.00 1.01943 *****
    25.00 1.01924
    26.00 1.01909
    27.00 1.01896
    28.00 0.83520 *****
    29.00 0.83459
    30.00 0.83395
    31.00 0.83331
    32.00 1.04453 *****
    33.00 1.04468

This occurred for simulations of 4 different systems. In addition,
running on the Amber 10 version of PMEMD, this problem disappears.
While that's a solution for future simulations, I wonder if I can
salvage any of the data in these trajectories. Can I simply use every
4th snapshot for analysis, for example? Any ideas would be greatly
appreciated.

Thanks;
-Sergio

_______________________________________________
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