[gmx-users] pull code geometry direction

[Fabian Knoch]

Dear users,

I got some questions concerning the pull code (Gromacs 5.1.2) more 
precisely about the constant force protocol. I searched already quite a 
bit but couldn't find any satisfying answers.

If I understand it correctly the two major options one has to set for 
the pull-geometry is "distance" or "direction".

For "distance" I have to specify two groups and Gromacs calculates the 
vector connecting both groups and applies the force along the connecting 
vector.  Hence, the simulation box has to be more than twice as large as 
the length of the connecting vector because otherwise the vector cannot 
be uniquely determined (minimum image convention).

For the second option "direction" one has to specify a vector 
"pull-coord1-vec " along the force will be applied.

My first question is what's the difference between option "distance"  
with "pull-coord1-dim =  N N Y" and "direction" with "pull-coord1-vec = 
0 0 1" ? If I understand the mainual correctly, these two options should 
give identical results.

Now concerning my problem:

For my pulling simulations (pulling of a protein) I cannot choose 
"distance"  because then the box size becomes too large and I simply 
have too much water in it.

So I need to take the "direction" option with "pull-coord1-vec = 0 0 1" 
(pulling in z-direction) and further introduce a position restraint 
(C_alpha N-terminus) ensuring that the protein is not floating around 
too much.

Now after I run the simulation for a bit, gromacs returns an error 
saying that the distance is larger than 0.49 the box size and I should 
consider using "pull-geometry = direction-periodic".

Could anyone explain to me why I receive this error message. I don't 
understand what distance gromacs is referring to. I assume the distance 
between the pulling group and the point (0,0,0), though. Second, I don't 
understand the distance reference in general as I specified 
"pull-geometry =direction". For my understanding the box size should not 
matter as a constant force (with fixed direction) is simply applied to a 
reference group.  It is conceptually identical to a constant electric 
field (in fixed direction) which acts only on a reference group. It 
should only be important to make the box large enough to avoid 
self-interactions.

To bypass this error I have to set "direction-periodic" but then I 
cannot use isotropic pressure coupling anymore and thus have to use 
semiisotropic pressure coupling.

So my questions sum up to why is not possible to use pull-geometry = 
direction with a box size protein length + say 4 nm and isotropic 
pressure coupling ?

Attached are two pull-code snippets from my .mdp file:

(i) "pull-geometry = direction" ----> Not working

pull = yes
pull_ngroups = 1
pull_ncoords = 1
pull-coord1-groups = 0 1
pull_group1-name = pull-atom
pull-coord1-type = constant-force
pull-coord1-geometry = direction
pull-coord1-vec = 0.0 0.0 1.0
pull-coord1-k = -300

(ii) "pull-geometry = direction-periodic" ---> working

pull = yes
pull_ngroups = 1
pull_ncoords = 1
pull-coord1-groups =0 1

pull_group1-name = end-atom
pull-coord1-type = constant-force

pull-coord1-geometry = direction-periodic
pull-coord1-vec = 0.0 0.0 1.0
pull-coord1-k = -300

pcoupltype          = semiisotropic
tau_p                   = 2.0
ref_p                    = 1.0 0
compressibility    = 4.5e-5 0


Thank you

Fabian