[gmx-users] Re: PMF profile using g_wham with pull_geometry = distance for ion channel

[chris.neale at utoronto.ca]

Dear Tom:

Your idea will work, but only for the one replica centered exactly at  
0.0; the other ones will need to be rerun. But you really should not  
need to rerun very many replicas, unless I misunderstand your setup.

The problem is this: if your center of restraint is close enough to  
0.0 that sampling can jump from one side of 0.0 to the other and you  
essentially sampled a bimodal population. E.g. your distance was set  
to 0.1 and your sampling starts at 0.1, but migrates to negative  
values... the pulling is now to a different center of restraint and  
hence you are out of equilibrium. This means that the sampling itself  
is useless.

For the case exactly restrained to 0.0, you have 2 options. Option 1:  
can you simply write a script to look at the -pf output and adjust the  
sign of the -px output based on the sign of the force?  Option 2: If  
you have the .xtc file, you could analyze distances with whatever tool  
you like (indeed, your suggestion of mdrun rerun seems to be valid  
unless it crashes...). Just hack the output from whatever  
mdrun/g_dist/etc. tool you use into the correct format for g_wham  
input. I realize that your .xtc will have been saved far less  
frequently than your -px output, but usually the addition data that  
you will lose here is highly correlated and your PMF will still  
contain the important data -- test this for yourself by extracting  
every 10th line from your -px file as below and rerun g_wham on that  
section:

cat my.xvg | sed '-n 1~10p' > new.xvg

IMPORTANT NOTE: Both of these methods will work for two cases and not  
work for a third. They will work at:

1. center = 0.0
2. center is "far enough" away from 0.0 (see below).

They will fail at:

3. As mentioned above: Your center of restraint is close enough to 0.0  
that sampling can jump from one side of 0.0 to the other and you  
essentially sampled a bimodal population.

Chris.

-- original message --

Hi again,

Just before I reside myself to rerunning a large number of simulations I
was wondering if it is possible to just rerun the simulations where the
ion is above and below the reference using pull_geometry=position. Then
could I combine these with the other simulations where
pull_geometry=distance to get the complete PMF?

Tom