Hello,
On Mon, Aug 8, 2011 at 10:09 PM, setyanto md <stwahyudi.md.gmail.com> wrote:
> Dear Amber User and developer,
>
> I read at AMBER 10 Manual, page 37, at the desription of "cut", it is said
> that "For PME, the cutoff is used to limit direct space sum, and the
> default
> value of 8.0 is usually a good
> value. When igb>0, the cutoff is used to truncate nonbonded pairs (on an
> atom-by-atom basis); here a larger value than the default is generally
> required."
>
> My simulation use explicit solvent mode, so this mean igb=0 by default,
> according to the manual it is enough using cutoff 8 angstrom.
>
Typically, yes. You will hear some disagreement about it, but for explicit
solvent using PME, the cutoff only *really* applies to the VDW interactions,
which decay as 1/r^6.
While it does also limit the direct space sum in PME, the reciprocal space
sum picks up with the direct space sum doesn't get, so you get the full
electrostatic interaction. The direct space scales as N^2 and the
reciprocal space sum scales as N log N, so a smaller cut yields a cheaper
calculation (to an extent).
For GB, however, the whole electrostatic calculation is a direct sum over
interacting pairs with no reciprocal space sum to get the rest of the
interactions. The issue with a small cutoff here is that electrostatic
interactions decay as 1/r, which is actually a divergent series, so it's a
very long-range force. As a result, you'll need a much larger (perhaps even
infinite) cutoff to get a decent picture of what's happening.
> I want to make sure if this cutoff only related with explicit and implicit
> solvent mode ?
>
I don't know what you mean here. It's a general nonbonded interaction
cutoff...
>
> How about th size molecule and number of amino-acid ? is this also need to
> considerate or not ?
>
Not typically. If the cutoff is significantly smaller than the system size,
then it doesn't really matter how big the system is, you're going to see the
above mentioned effects anyway.
> My molecule have 388 amino acid and the size is 53.04 x 55.07 x 58.13 A
> (volume total is 169792.66)
>
> questions:
> 1. Is this enough to use 8 angstrom as cutoff or need larger cutoff ?
>
8 Angstrom cutoff would probably be enough for an explicit solvent
calculation.
> 2. What is the significantly different between 8 and 12 cutoff ?
Computational cost, and a *very* little bit of accuracy in the VDW
calculation. A 12 Å cutoff will be more expensive, since the direct space
sum is larger.
However, the direct space sum does parallelize better than the reciprocal
space sum, so you will probably see better scaling (but worse overall
performance) using a larger cutoff.
HTH,
Jason
--
Jason M. Swails
Quantum Theory Project,
University of Florida
Ph.D. Candidate
352-392-4032
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Mon Aug 08 2011 - 20:00:03 PDT
