AMBER Archive (2005)

Subject: Re: AMBER: nscm

From: David A. Case (case_at_scripps.edu)
Date: Tue Apr 12 2005 - 10:55:58 CDT


On Tue, Apr 12, 2005, Ilyas Yildirim wrote:
>
> I am doing a free energy difference calculation from ethane -> methane in
> gas phase. This is a test case. I created the structures in Xleap and did
> 202 dv/dl simulations for 100 ps.; 101 for nscm=200 and 101 for nscm=50.
> Then I used the trapezoidal rule to calculate the integral and here are
> the results:
>
> for nscm=50 : Integral = -0.157530606498188
> for nscm=200: Integral = -0.199276122465974
>
> Is this something normal to expect when we set the number of steps for
> center of mass regulation to different numbers? I would have expected no
> change in the results. Did anybody encounter with a similar case? Thanks
> in advance.

If you do two simulations that are different in any respect (even on different
machines), you should expect small differences, due to the nature of molecular
dynamics. Round-off, for example, is different after you remove translation
and rotation.

So getting a free energy that is different by 0.04 kcal/mol following 20 nsec
of simulation is not at all surprising (to me, anyway).

...dac

-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber_at_scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu