[gmx-users] Multiple Simulations

[David]

On Mon, 2004-01-12 at 16:51, Anton Feenstra wrote:
> David van der Spoel wrote:
> > Doesn't that depend on the kind of properties? If the simulations are 
> > under the same conditions (T,P etc.) then I'd say one can just 
> > linearly average most properties.
> 
> That depends on the starting conditions, doesn't it? If I would simulate
> a protein starting 10 times from an x-ray structure and 10 times from a
> rough homology model (or, worse, a high-T unfolded structure), I wouldn't
> trust the linear averages...
> 
> In other words, the x-ray structures are probably of a lower (free) energy,
> making therefore the whole trajectory more 'likely', than those started
> from the high (free) energy state. It could come down to assigning (free)
> energies to the independent trajectories, which may be as simple as taking
> the average potential energy over the trajectory. But I'd like to know if
> anyone has done some kind of (rigorous) comparison on methods to do this?

Aha, I somehow (naively) though you meant equilibrium trajectories... My
experience is that the bad structures have higher energy, not rigorous
though. Have you seen the recent paper from the Alan Mark group on
comparison of NMR and Xray structures, all simulated for a long time?
Not surprisingly, the Xray structures are more stable (so indeed, forget
about homology models, even though I have just published a paper on
simulations of those... [J. Comp. Aid. Mol. Des. 17, p 551 (2003)]).

-- 
David.
________________________________________________________________________
David van der Spoel, PhD, Assist. Prof., Molecular Biophysics group,
Dept. of Cell and Molecular Biology, Uppsala University.
Husargatan 3, Box 596,  	75124 Uppsala, Sweden
phone:	46 18 471 4205		fax: 46 18 511 755
spoel at xray.bmc.uu.se	spoel at gromacs.org   http://xray.bmc.uu.se/~spoel
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++