[gmx-users] long simulation

[David van der Spoel]

On Sun, 26 May 2002, Dmitry Kovalsky wrote:

>sorry if a question became trivial but my experience is not too big to 
>reply on the question by myself. I've been reading mailing list archive 
>concerning PME and RF and steel cann't make decision about mdp parameters to 
>chhose for my system. 
>I want to simulate my flexible protein that has a charge +4 in water in long 
>timescale (near 50 ns). And I want to choose most reasonable parameters for 
>my simulation. 
>
>For the correct simulation I have to insert one Na+ ion that binds into the 
>structure. So for neutral system it should be 1Na+ and 5 Cl- ions in the 
>system.
>In order to mimic native-like enviromnent I would like to insert 15 NaCl ion 
>pairs more, into my system (if it is correct, please advice me). Finally I 
>will compute it on 4CPU cluster (2x2CPU-node). 
First off, compute the salt concentration for you system, it may be high,
but it depends on the amount of water of course. 

>The question is that what the PME or the Reaction Field will be reasonable in 
>these conditions?
Opinions diverge here. RF is in principle for homogeneous systems, which
you don't have. PME is in principle for crystals which you don't have
either. All in all, PME will give you the best description of Coulomb
interactions, but it is more time consuming. Also, PME does not scale very
well over many processors, unless you have fast network (Myrinet/SCALI).
Test it with a 10 ps simulation. RF should not be a problem in scaling.

Concerning your parameters, when you use shake you can afford a timestep
of 2 fs.

Groeten, David.
________________________________________________________________________
Dr. David van der Spoel, 	Biomedical center, Dept. of Biochemistry
Husargatan 3, Box 576,  	75123 Uppsala, Sweden
phone:	46 18 471 4205		fax: 46 18 511 755
spoel at xray.bmc.uu.se	spoel at gromacs.org   http://zorn.bmc.uu.se/~spoel
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++