Structural inconsistencies, at least partially related to traj length

[Jason Held]

Hi folks,

I'm am running into an issue with a homopentameric protein plus membrane system (GROMACS) that I can't troubleshoot, and would appreciate any thoughts on how to fix this or if you've seen this issue before. I observe it using the newest MDAnalysis version 2.6.1, but I also observed it in version ~2.2 before I upgraded hoping it'd fix the problem, but it didn't.

The issue(s) are that I am seeing residues are in quite different positions in when viewing the same frame in NGLview versus writing the frame to a PDB, and there is also a difference in the structure depending on how many trajectories are loaded. For example when I load a single trajectory (1e6 steps/traj), the first frame looks correct based on my knowledge of the starting structure using NGLview. This is the center picture below, with a 'symmetrical' orientation of 5 important methionines highlighted in red at the center in an ion channel.

But, I am seeing two inconsistencies that worry me. 

1: if I save this same exact frame as a pdb with MDAnalysis using 'write', I get a 'wonky' orientation when I open it in chimera (image on the left), with a single red Met that is no longer symmetrical whatsoever. So, I am concerned by NGLview looks different than the pdb for the exact same frame.

2: I also get this same 'wonky', orientation of 1 methionine if I load up and concatenate 29 additional continuation trajectories after the first one using NGLview, which is the image on the right. So, I am concerned why the same frame looks different in NGLview depending on how many trajectories are loaded. If I 'write' this frame to PDB it also looks also wonky in Chimera

I cannot fix this issue using on-the-fly transformations like unwrap, it occurs whether or not I use a pdf, psf, or gro file as the starting structure for the universe, and it happens whether or not I try to adjust the pbc using GROMAC's trjconvert prior to making a universe in MDAnalysis.

And, there is one last wrinkle that is very confusing to me... If I export the 'single' trajectory (with a wonky pdb export but correct NGLView), or the concatenation of 30 trajectories (both wonky PDB and wonky NGLViews) from MDAnalysis and then load it with mdtraj, the atoms/structure traj look totally correct in mdtraj. And, mdtraj will export a pdb that also looks correct. In short, MDAnalysis doesn't appear to corrupting things, but I am sure having a hard time believing that based on the analyses above and am concerned about moving forward with it.

So, I'm looking for ideas about why MDAnalysis is consistently exporting a wonky PDB structure, and about the inconsistencies I observe with the same frame with NGLviewer depending on traj length, and what anyone might suggest to troubleshoot this since I really like MDAnalysis for analysis, but I want to be confident the underlying data is accurate.

Cheers,

Jason

[Hugo Macdermott-Opeskin]

Hi Jason, 

So we can help you better would you be able to share your code that you are using to  save your pdb's and load using NGLView?

Cheers, 

Hugo

[Jason Held]

Happy to, and if this doesn't spark any ideas I can make a shorter traj suitable for uploading to see if it happens on another setup.

To make a PDB of the first frame, I do this using write:

    u.trajectory[0]
    selectAtoms = u.select_atoms("all")
    selectAtoms.write(str(name + "_allAtoms_firstFrame.pdb"))

To view with NGL viewer i do this:

    view = nv.show_mdanalysis(u.select_atoms('protein'))
    view.background = 'black'
    view.camera = 'orthographic'
    view.add_ball_and_stick("296", color='red')view

[Jason Held]

Sorry, that last 'view' should be alone on the last line, but there was a formatting hiccup.