Re: [AMBER] Different Versions of Amber: Comparable results?

From: Jason Swails <>
Date: Tue, 11 Jan 2011 15:20:38 -0500


On Tue, Jan 11, 2011 at 3:05 PM, david condon <> wrote:

> Hey all,
> I have 3 simple questions (tried searching for this already and asked
> around, no one is absolutely sure about this):
> 1) Are results from pmemd and sander comparable? I have stopped a
> simulation that was running on SANDER and continued it using PMEMD.
> As far as I know, the only substantial difference is in how its
> processed, and there is less information generated from PMEMD but
> that's not important for what I'm doing.

They should be the same to within machine precision. The problem with the
disclaimer is that machine precision is enough to make 2 same starting
snapshots end in 2 completely different places after enough time has
passed. This is an issue only for perfect reproducibility, not in the
actual science/method.

> 2) Is it acceptable to run a single simulation with different versions
> of AMBER (e.g. at 200 ns switch from Amber 9 to Amber 11)?

Sure. The results should not change from one version to another for the
methods that are contained in both (unless a bug has been fixed and your
older version is not up-to-date with bugfixes)

> 3) Is it acceptable to switch from SANDER to PMEMD in the middle of a
> simulation? I did this a few days ago, and it looks fine. I just
> would like to be sure.

Again, sure. Look at the answer to your first question.

Hope this helps,

> Thanks for your time!
> -Dave
> _______________________________________________
> AMBER mailing list

Jason M. Swails
Quantum Theory Project,
University of Florida
Ph.D. Graduate Student
AMBER mailing list
Received on Tue Jan 11 2011 - 12:30:04 PST
Custom Search