Re: AMBER: Maximum system size in simulated annealing

From: Andreas Svrcek-Seiler <svrci.tbi.univie.ac.at>
Date: Thu, 3 Aug 2006 11:23:13 +0200 (CEST)

>
> I succeeded, however, the problem is that I cannot avoid rather high
> velocities (vlimit) being often of the order of tens (sometimes hundreds). I
> think it is caused by the large size of the overall system which I simulate
> - it consists of almost 180 thousands atoms (10 thousands of heavy, 8
> thousand of hydrogens within the crystal and the rest is water solvent). Is
> it reasonable to use Amber annealing for such a big system? Is the numerical
> procedure consistent with so many atoms?
...System size does not influence numerics (besides loss of
accuracy way beyond the decimal point). Whether applying simulated
annealing to your system makes sense rather depends on how good your
initial coordinates are. You probably won't obtain "the right"
structure within reasonable time and there's no guarantee you might get it
even given infinite CPU time.

>
>
> The system was, of course, very carefully minimized (belly and nonbelly,
> around 60 thousands steps) before annealing simulation.
...this doesn't tell you anything but that the structure is close to a
local minimum - you're pretty high up at the first local minimum on the
way down from the top of Mount Everest.
I have never done this, but the first thing I'd try is taking the solute
alone and a fast electrostatic approximation (e.g. diel=r), maybe apply
some restraints to fix atoms I believe to be close the "the right"
position, do simulated annealing on that and see whether (and how much)
the energy goes down after one or some cycle(s) of
heating-cooling-minimization.

One might also simply try to estimate E_VdW, E_elec and E_bond for a
"reasonable" system of "your" size and see whether your structure
has energies at least of the same order of magnitude.

You might also cut your system to smaller pieces and do some experiments
on these pieces.

...Again: this is what I'd try, but I have never tried it.
Experiment is the best teacher :-)

good luck
Andreas






-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
Received on Sun Aug 06 2006 - 06:07:09 PDT
Custom Search