[gmx-users] g_rmsf over time windows

[giordano mancini]

Hi all,

I have the following problem: I have a simulation
of 10 ns of a protein in water. I have extracted the protein trajectory
with trjconv -pbc nojump and the fitted it to a reference structure
with trjconv -fit rot+trans.
If I run g_rmsf -nofit over two consecutive chunks of 5 ns i get very
similar rmsf values (so the fluctuations are are the same in the 2 time
windows). However if I run g_rmsf over the whole trajectory i get much
higher values of rmsf (up to 0.1 - 0.15 nm). The difference become 
higher if use 4 windows of 2.5 ns each (i.e. the rmsf over 2.5 ns are 
smaller).
I think there's something wrong with the reference structure which is
different in the various cases so that with shorter trajectories the std
dev is smaller; however I always use the same .gro (or tpr). May be are
the -b and -e flags?
Use of tpr/gro reference, fit with g_rmsf, calpha or backbone,
or -res option does not change things.
I have tried with gromacs 3.3.1 and 3.3.3

Here 's my complete syntax:

$> trjconv -s box.gro -f nojump.xtc -n index.ndx -o nj -pbc nojump
$> trjconv -s box.gro -f nj.xtc -n index -fit rot+trans -o fitted
$> g_rmsf -s box.gro -f fitted -n index -o rmsf$start -b $start -e $end
for chuncks

$> g_rmsf -s box.gro -f fitted -n index -o rmsf

thanks for your help
giordnao mancini