XTC, GRO and PDB data file do not correctly reflect triclinic basis.

[Miro-Astore]

Expected behavior

The example files XTC PDB and GRO files have triclinic bases (angles 60 60 90)
testsuite/MDAnalysisTests/data/adk_oplsaa.xtc
testsuite/MDAnalysisTests/data/adk_oplsaa.pdb
testsuite/MDAnalysisTests/data/adk_oplsaa.gro

Actual behavior

Visually they appear to have a monoclinic basis (angles 60 90 90) according to the definitions https://en.wikipedia.org/wiki/Triclinic_crystal_system
This caused me some confusion when working on #3409

Perhaps they should be changed to reflect their metadata?

Code to reproduce the behavior

import MDAnalysis as mda
from MDAnalysis.tests.datafiles import PSF, DCD,  GRO, PDB, TPR, XTC, TRR,  PRMncdf, NCDF

u = mda.Universe(GRO,XTC)
for ts in u.trajectory:
    print(u.dimensions)
u = mda.Universe(PDB)
print(u.dimensions)

....

Current version of MDAnalysis

• Which version are you using? 2.0.0-dev0
• Which version of Python (python -V)? 3.7
• Which operating system? Linux

[richardjgowers]

I think the issue here is that we use a loose definition of triclinic, where it means at least one corner isn't 90, rather than all three aren't 90. This is because under the hood the code is split into either orthorhombic or not (what we call triclinic) code paths.

[Miro-Astore]

I think the atoms are not arranged in the correct unit cell.

In VMD try this:
mol new adk_oplsaa.gro
package require pbctools
pbc box_draw

You'll see that the true unit cell doesn't align with the monoclinic box which the atoms are placed in. I'm not sure why this is the case but it will cause issues when creating the virtual images.

I could fix the GRO PDB XTC and TRR files but TPR would be harder to fix.

[IAlibay]

I might be misunderstanding the issue, but I've always been under the impression that these files were not synthetic but instead gromacs generated. Hence, I'm not sure that they should be fixed (at least not manually).

They far predate my time with MDA, so maybe @orbeckst might have more information?

[Miro-Astore]

That's why I was confused. The trajectories seem to have been wrapped to a monoclinic box with angles 60 90 90 rather than the angles indicated by their metadata. 60 60 90.

Current versions of gromacs put triclinic simulations in an orthogonal box and then you need to use trjconv -ur trilinic to see the 'true' unit cell. I'm not sure if this is just a version issue. In any case. I'm not sure these files will be producing pbc images which tessellate in their natural basis. I'm not sure if that's a known issue.

[lilyminium]

Current versions of gromacs put triclinic simulations in an orthogonal box and then you need to use trjconv -ur trilinic to see the 'true' unit cell. In any case. I'm not sure these files will be producing pbc images which tessellate in their natural basis.

Maybe someone selected the wrong output box cell? It does look better if we wrap the atom coordinates:

import MDAnalysis as mda
u = mda.Universe("adk_oplsaa.gro")
u.atoms.wrap(inplace=True
u.atoms.write("wrapped.gro")

[orbeckst]

These files were GROMACS generated a long time ago.

As @richardjgowers #3410 (comment) said: Any angle != 90º and we call it "triclinic".

[orbeckst]

That's why I was confused. The trajectories seem to have been wrapped to a monoclinic box with angles 60 90 90 rather than the angles indicated by their metadata. 60 60 90.

What meta data are you referring to?

[Miro-Astore]

Perhaps meta data is not the correct term but I am talking about the cell basis vectors in the gro, PDB, TPR and XTC files. The xyz coordinates of the atoms reflect a basis with angles 60 90 90 between vectors while the pbc box should be 60 60 90. See attached screenshot and @lilyminium 's

pbc box_draw was used in VMD to show the virtual pbc cell. You can use the attached vmd state vector to check this visualization. This is a minor issue but I just noticed it while I was working on the mentioned pull request.
test.vmd.txt