[gmx-users] criterion for good sampling

[David van der Spoel]

On Tue, 30 Apr 2002, Y. U. Sasidhar wrote:

>I s there any rule of thumb to ascertain whether conformational sampling 
>in a MD run is adequate ?
>One review article ( by van Gunsteren & Berendsen ) seems to suggest that a
>physical property independent of simulation length is a necessary but not
>sufficient condition.
Basically you will never sample the complete configuration space in any 
simulation. But: for simple systems like pure liquids you can look at 
things like diffusion, dielectric constant, and see if you get the same
value for parts of the trajectory (i.e. cut the traj in 2 - 5 pieces and 
compute these properties). If the results are similar, you have adequate 
sampling.

For complex systems like proteins, you will only sample one or a few 
potential minima at best. Your best bet is to run every simulation at 
least twice (with different starting conformation and energies), and then 
compare the trajectories, e.g. with g_mdmat, essential dynamics analyses
and so on. Keep in mind that you have to convince a crystallographer who 
doesn't believe in modelling that your results are meaningful.

-- 
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
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++