[gmx-users] using the pull code to constrain the COM of every molecule in a system

[Nicholas J. Dunn]

Hello all,
I have what may seem like a strange question: Is it possible to use the pull
code in gromacs 5 to run a simulation where the center of mass of every
molecule in the system is constrained?  I'm trying to set up such a simulation
so that I can sample from the set of atomistic configurations that map to a
particular coarse-grain configuration where the coarse-grain sites are defined
as the COM of the molecule.  I'm starting with a proof-of-concept simulation of
a system with two generic diatomic molecules (with a harmonic bond), each
constrained to their respective center of mass.  Based on my reading of the mdp
options, I've put the following settings into my .mdp file:
    ; Pulling code    pull                = constraint    pull-geometry       =
distance     pull-dim            = Y Y Y    pull-constr-tol     = 1e-6   
pull-ngroups        = 2    pull-ncoords        = 2    pull-group1-name    = LJ1
   pull-group2-name    = LJ2    pull-coord1-origin  = 1.2 1.2 1.2   
pull-coord2-origin  = 0.1 0.1 0.1     pull-coord1-groups  = 0 1   
pull-coord2-groups  = 0 2
However, when I run this simulation I get an error as if the system is blowing
up: 
    Program gmx, VERSION 5.0    Source code file:
/gpfs/home/njd166/work/src/outside_code/gromacs/gromacs-5.0/src/gromacs/pulling/pull.c, line: 541
    Fatal error:    The pull constraint reference distance for group 0 is <= 0 (0.000000)
If I change the pull setting to "umbrella" and define a spring constant for each group, the simulation runs as expected - each of the molecules are tethered to their reference point.  Is there some fundamental reason why using constraints isn't possible, or have I made an error something in the settings?
Thanks,- Nick Dunn