[gmx-users] Unexpected trjconv -nojump behaviour

[Tsjerk Wassenaar]

Hi Trayder,

> The box is identical (copy+pasted), the orientation varies no more than
one
> would expect frame to frame (membrane protein). There is a 20 A jump
> between the first and second frame but shouldn't -nojump still be keeping
> the protein whole?

The idea of what jumped may change quite a bit if you have a 2nm shift.
Also, nojump only keeps things whole if the molecules are whole up to PBC
shifts. And they aren't as you find from your output.

> And shouldn't -pbc mol be keeping my "molecule" together? Even if its
doing
> it by fragments, I can't reason why it would take a fragment almost
> completely inside the box (<1% atoms outside) and place it on the opposite
> side of the system.

The routine takes the first atom and then sees whether following atoms need
to be moved over.

> I realise the system is trickier than most because its a rhombic
> dodecahedron with the protein fitting quite tightly, but I've run dozens
of
> batches of simulations and only 2 are causing me any trouble. I can only
> assume I'm making some invalid assumptions about how the commands work.

Probably. The routines are quite clear (and simple, and bugfree). This
usually means there is a mismatch somewhere, between reference and
trajectory or between successive frames.

Try these two things:

Visualize the .tpr and the first frames with the box and see what unjumping
could do.

Extract the original first frame from the trajectory and make that whole,
then use that as reference with -pbc nojump.

Cheers,

Tsjerk