[gmx-users] how to choose -beginfit for g_msd?

[Shang-Te Danny Hsu]

I am trying to determine lateral diffusion coefficients of molecules in 
lipid membranes using g_msd. The trajectory is 20 ns long and I'd like 
to analyse the segment of 5-20 ns with 2.5 ns intervals for statistics. 
The command used is as follows:

g_msd -f A.xtc -s A.tpr -n group.ndx -b 5000 -e 20000 -trestart 2500 
-lateral Z (with varied -beginfit values)

I noticed that the output depends very much on the choice of -beginfit 
(see examples below) and I wonder whether -beginfit is an absloute value 
like -b and -e or it is a relative value with respect to the time point 
defined by -b? I assume the later is the case because the diffusion 
coefficients are zero when -beginfit (7.5 ns) is set to be half of the 
length of trajectory segment (5-20 ns).

-beginfit   0

D[         A] 0.072 (+/- 0.122) 1e-5 cm^2/s
D[         B] 0.061 (+/- 0.096) 1e-5 cm^2/s

-beginfit   3000

D[         A] 0.052 (+/- 0.117) 1e-5 cm^2/s
D[         B] 0.045 (+/- 0.082) 1e-5 cm^2/s

-beginfit   5000

D[         A] 0.028 (+/- 0.040) 1e-5 cm^2/s
D[         B] 0.025 (+/- 0.004) 1e-5 cm^2/s

-beginfit   7500

D[         A] -0.000 (+/- 0.188) 1e-5 cm^2/s
D[         B] 0.007 (+/- 0.203) 1e-5 cm^2/s

If so, when the standard deviation derived with six restarting time 
points is twice as large as the average value, how and when are we 
confident that the simulation is long enough?

Any suggestion?

-- 
Shang-Te Danny HSU
Department of NMR Spectroscopy
Bijvoet Center for Biomolecular Research
Utrecht University
Padualaan 8, 3584 CH Utrecht, the Netherlands
phone: +31-30-2539931 | fax: +31-30-2537623
e-mail: hsu at nmr.chem.uu.nl