Lubos -
I think it is well known that sander 6 and sander 7 produce slightly
different results due to differences in the step integrator. Less well
known are differences due to changes in sander 7 regarding how molecular and
atomic virials are calculated. Typically, this difference is only seen if
you have intermolecular crosslinks, and the difference arises based on what
6 vs. 7 consider to be a molecule. PMEMD 3.1 is a hybrid of 6 and 7 in that
it always handles virials like sander 6; the changes in virials handling in
7 were associated with extra points code, and I did not attempt to
incorporate this stuff into pmemd. So, the difference in results between
pmemd in amber 6 mode vs. pmemd in amber 7 mode is mostly due to the amber 7
step integrator change. PMEMD in nocit mode should be fairly close to
sander 6 in all respects. So I think you are fine. Glad you got it
working.
Regards - Bob
----- Original Message -----
From: "Lubos Vrbka" <shnek.tiscali.cz>
To: <amber.scripps.edu>
Sent: Wednesday, April 14, 2004 3:21 PM
Subject: Re: AMBER: non-isotropic pressure scaling
> hi guys,
>
> here are the final results and a short summary. test was performed on
> dhfr benchmark, 100 steps, printout every single step. ntb was changed
> to 2 (constant pressure) and ntp to 2 (nonisotropic scaling)
>
> - sander7 and pmemd-3.1 produce exactly the same numbers when
> amber7_compat = 1 is used
>
> - pmemd-3.1 results in nocit or amber 6 cit mode are the same. when
> compared to sander7 results there are some differences:
>
> sander 7 averages for 100 steps:
> ================================
> NSTEP = 100 TIME(PS) = 510.150 TEMP(K) = 299.05 PRESS = -148.5
> Etot = -57762.4772 EKtot = 14005.7262 EPtot = -71768.2034
> BOND = 476.0346 ANGLE = 1266.5567 DIHED = 987.0423
> 1-4 NB = 537.4913 1-4 EEL = 6646.8490 VDWAALS = 8324.2655
> EELEC = -90006.4428 EHBOND = 0.0000 RESTRAINT = 0.0000
> EKCMT = 6124.0382 VIRIAL = 6864.3012 VOLUME = 230807.7450
> Density = 1.0139
> Ewald error estimate: 0.7657E-04
>
> pmemd 3.1 cit (amber 6) averages for 100 steps:
> ===============================================
> NSTEP = 100 TIME(PS) = 510.150 TEMP(K) = 299.73 PRESS = -167.65
> Etot = -57738.9175 EKtot = 14038.5819 EPtot = -71777.4994
> BOND = 474.3691 ANGLE = 1264.2984 DIHED = 987.3340
> 1-4 NB = 539.1834 1-4 EEL = 6650.9716 VDWAALS = 8340.4301
> EELEC = -90034.0861 EHBOND = 0.0000 CONSTRAINT = 0.0000
> EKCMT = 6125.7684 VIRIAL = 6961.3817 VOLUME = 230726.5228
> Density = 1.0143
> Ewald error estimate: 0.7775E-04
>
> the question is whether the differences significant and whether they
> will vanish or propagate for longer md runs.
>
> however, when amber7_compat = 1 is used, nonisotropic pressure scaling
> (ntp=2) SEEMS to work fine.
>
> the only required change to the pmemd-3.1 source is commenting out
> appropriate part of the mdin_ctrl_dat.f90 file where input is tested for
> NTP .eq. 2 (approximately lines 388-393 of that file). all other files
> already contain the necessary routines (you can check runmd*.f90)
>
> thanks to all people who helped me when solving this problem. regards,
>
> --
> Lubos
> _._"
> -----------------------------------------------------------------------
> The AMBER Mail Reflector
> To post, send mail to amber.scripps.edu
> To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
>
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
Received on Sun Apr 18 2004 - 14:53:00 PDT
