Pulling in non-rectangular box

GROMACS version: 2020.3, 2020.0, 2018.8
GROMACS modification: No
Dear Gromacs Users,

I encountered an issue when trying to use constant velocity pulling with the gmx pull code with a membrane system with a hexagonal shape.
Initial structure/system:
Transmembrane protein in hexagonal membrane patch, 136826 atoms, centered so that k97 (pull-group1) is centered in the box with trjconv -ur compact -pbc mol -center. This should also put the center of mass of the membrane close to the center of the box.
Dimensions (last line of .gro file): 11.26904 9.75928 12.10227 0.00000 0.00000 -5.63452 0.00000 0.00000 0.00000

What I try to do:
Pull k97 in x direction (ultimately I want to pull in all three directions, but only the x direction causes problems) while holding the membrane stable.

pull = yes
pull-ncoords = 1
pull-ngroups = 2
pull-group1-name = k97 ;; specific lysine that I want to pull at
pull-group2-name = MEMB ;; whole membrane
pull-group2-pbcatom = 25084 ;; arbitrary membrane atom relatively close to the center of the box
pull-pbc-ref-prev-step-com = yes
pull-coord1-groups = 1 2
pull-coord1-type = umbrella
pull-coord1-geometry = direction
pull-coord1-vec = -1 0 0
pull-coord1-dim = Y N N
pull-coord1-rate = 0.3 ;; fast, just for testing
pull-coord1-k = 10 ;; kJ mol^-1 nm^-2
pull-coord1-start = no

gmx grompp does not complain about this, but when I start my simulation, gmx mdrun crashes at the first step with this error message:
“Fatal error:
Distance between pull groups 1 and 2 (4.842380 nm) is larger than 0.49 times
the box size (4.782055).
You might want to consider using “pull-geometry = direction-periodic” instead.”

The initial distance should not be that large and also the x box size mdrun returns is much smaller than the one in the gro file.

Am I overseeing anything? Has anyone encountered similar issues?

Many thanks,

what about to use pull-coord-geometry = cylinder?
That is what works best for pulling a lipid from a membrane layer
(see http://manual.gromacs.org/current/reference-manual/special/pulling.html?highlight=definition%20center%20mass).

Best regard

Hi Alessandra,

thank you very much for your suggestion!
I only got around to testing it now - unfortunately the cylindrical geometry doesn’t work either…