[gmx-users] umbrella sampling pull.pdo output

[Manohar Murthi]

hi everyone

i'm trying to calculate the free energy of binding of
a guest and host in solution by umbrella sampling.

my pull groups consist of individual atoms on the host
and guest. i tried a force constant of 1/100*single
bond strength.

when the umbrella distance between the host and guest
atoms is set to {0.4,0.4,0.4} (i.e. .4 nm separation
in x,y, &z), i notice that the pull.pdo file
eventually shows the separations in x,y,&z becoming
very small - on the order of 0.02nm, which should
surely result in an energy explosion. however, a 100ps
simulation runs to completion with no complaints.
examining the distance between the 2 pulled atoms
using g_bond shows that they in fact remain around
0.7nm (=sqrt(3*.4^2)) apart, as expected. visual
inspection of the trajectory confirms that the pulled
atoms don't approach each other too closely.

is this perchance a bug or am i doing something wrong?

i start the simulation from a random configuration
selected from a long unconstrained nvt run using

grompp -f umbrella.mdp -o u4 -t nvt1.trr -time 100 -n
index.ndx (where index.ndx defines energy groups),

then run using
mdrun -s u4.trr -pi pull.ppa -pd u4 -pn pull.ndx.

my pull.ppa file looks like this:
verbose = yes
runtype = umbrella
reference_group = zn152
group1 = n204
reftype = com
k1 = 4393.2
pos1 = 0.4 0.4 0.4

and pull.ndx is:

[ zn152 ]
152

[ n204 ]
204

please let me know if more information is required to
understand the problem.

any suggestions would be greatly appreciated.

thanks
mo

ps - i also get an error running g_bond:
Fatal error:
No distribution....(i0 = 999, i1 = 1)? ? ! ! ? !

with no histogram being output. i thought this may be
to do with the fact that the atoms aren't bonded, but
have yet to check on this.

__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com