AMBER Archive (2003)

Subject: Re: PME cutoff box size

From: darden (darden_at_gamera.niehs.nih.gov)
Date: Tue Mar 25 2003 - 13:49:57 CST

Suzie
Actually its not the cutoff that matters so strongly as
the distance to the face. That is, PME energies and forces
are independent of cutoff (the reciprocal parameters grid size
and spline order along with ewald parameter beta need to adjust
with the cutoff however). So, if it weren't for the van der Waals
terms which are also cutoff you could run pme with a cutoff of
4A and appropriate reciprocal parameters and do fine. On the other hand
you could use a big cutoff of say 12A in your case and do fine. The
one caveat here is that the cutoff must be less that 1/2 the smallest box
dimension. This is due to amber's use of the minimum distance criterion,
which means that you can only calculate minimum image pairs; i.e. pairs
that are within 1/2 the smallest dimension away (periodically wrapped if
necessary).
So the real "law" is that you need plenty of water to screen undesired
periodicity artifacts. That is you don't want much interaction with your
own periodic images. To understand this latter think of a continuum
treatment where your protein is surrounded by a dielectric of 80, so (at
least at a distance) electrostatic interactions go like
E = 332*q1*q2 / (80*r12) or E ~ 4*q1*q2/r12 where q1 and q2 are partial
charges, r12 is distance between in angstroms and energy is in kcals.
So you would like plenty of distance (e.g. 20A) between protein and its
periodic images. For a more quantitative view using Poisson-Boltzmann
equation with periodic boundary conditions, see Hunenberger and mccammon
Biophys Chem 78 pp69-88 April 1999
Hope this helps
Tom Darden

 On Tue, 25 Mar 2003 kbyun_at_wesleyan.edu wrote:

> Hi,
> Could someone explain to me the "law" about PME
> that the cutoff must be less than the distance of the
> solute from the face of the box? i.e. cutoff =9A,
> distance to box = 10A. I would really appreciate an
> explanation, references that point me to this.
>
> Thanks,
> Suzie Byun
>