[gmx-users] minimum image violation

[Mark Abraham]

On 21/06/2011 3:43 PM, Kavyashree M wrote:
> Sir,
>
>      I am extremely sorry for this question again :(
> but I wanted to know this violation exists only in
> first 50ns but after that even though there appears
> to be a point of violation its only 1.39nm which is
> o0.01nm less than the cut off which I hope does
> not cause serious trouble (as the minimum bond
> length is of the order of 1 Angs) So can I make use
> of this part of trajectory?

The purpose of the minimum image convention is to help model the real 
conditions under which your protein exists. The usual case is one of 
(effectively) infinite dilution. However, if you have a protein atom 
that is influenced by an atom of a periodic image of the protein, then 
you are not close to modeling infinite dilution. Arguably, a water atom 
that can see two different periodic images is also unrealistic, but 
probably this will be a lower-order effect, and vary from case to case.

So as soon as you get a violation, further data about your system is 
tainted from the previous existence of unrealistic conditions. Even just 
looking at the first part of the simulation before the violation 
(assuming it has equilibrated properly so far) is questionable, because 
you have imposed on its ensemble the restriction that it cannot grow 
larger than a given diameter for a given orientation. It's up to you to 
judge whether the effect is small enough to ignore, and that depends on 
lots of things we can't know.

There exist a large number of areas of biomolecular simulations where 
our inability to exhaustively sample ensembles has limited our ability 
to quantify the size of various effects. The effect of violations of 
minimum-image conventions on proteins is one of those areas.

The important lesson here is that doing a simulation blindly runs a 
large risk of wasting resources. Ongoing attention to lots of details is 
important. As in many fields of human endeavour, there are lots of 
lessons that have been written in the blood of unfortunate people before 
you. In research, there will be lessons written in your blood. The trick 
is to learn from the former in order to minimize the latter :-)

Mark