AMBER Archive (2005)

Subject: Re: AMBER: Dummy Atom Types on both states

From: yipinl_at_umich.edu
Date: Tue Feb 15 2005 - 09:31:56 CST


Hello,
I have a relevant question. Since dummy atoms can only be in one state (lambda =
1), if we want to estimate the error of free energy calculation, should we use
different starting conformations to redo the calculation, instead of doing in
the "forward" and "reverse" direction?
Thank you very much!
Yipin

Quoting Thomas Steinbrecher <steinbrt_at_scripps.edu>:

> Hello
>
> > Dummy atoms are just atoms (or atom TYPES). They have some parameters too,
> > except that the non-bonded parameters are set to be zero. Let's say that I
>
> The problem with dummy atoms does not lie in sander itself but is a
> statistical effect that makes your results unreliable. Dummy atoms (on
> both "sides" of the perturbation) can give rise to the origin singularity
> effect, see for example:
>
> T. Simonson, Mol. Physics, 1993, 80:2,441-447
>
> sander is designed to handle such problems by a modified potential mixing
> function (setting klambda>1) but that requires to have the dummy atoms
> only in one state, because the modification that prevents the origin
> singularity for dummies in the perturbed state would make it even
> bigger for dummies in the ground state.
>
> I have tried a code-modification to prevent this, but it is still work
> in progress.
>
> Kind Regards,
>
> Thomas Steinbrecher
>
> -----------------------------------------------------------------------
> The AMBER Mail Reflector
> To post, send mail to amber_at_scripps.edu
> To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu
>
>
>

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