alisterburt
commented
1 week ago Hi @rkjensen - this is really concerning, thanks for the report and for doing all of the correct controls. I can confirm that up until now we only made sure that RELION could refine stacks and that we could import and further refine results in M.
It's very much possible that our RELION output is off in a way that is internally consistent (tilts are properly related in Fourier space) but actually wrong (product of tilt matrix and particle matrix is wrong in the same way for all particles)
I'll have to dig into this properly - #21 will need to be done properly which is a bit of a can of worms...
cc @dtegunov this is a big blocker for release, stops people using priors on orientation
dtegunov
rkjensen
I've tried to import an old-style RELION star file (RELION <3.1) to extract particles (ribosomes) with WarpTools (if needed see context below). If I extract with the --3d flag, and run relion_reconstruct, I get a perfectly fine looking ribosome, that I can then use for a round if local refinment, which for around 1000 particles goes to 18 A. If I do the same with the --2d flag, the angles don't work and local alignment just gives me noise. Global refinement is still running, but is currently at 27 A, looking like a ribosome, so I think the subtilt stacks as such are fine, and that it is rather an issue with the euler angle conversion. Let me know if you are interested in any of the star files.
EDIT: The 3d subtomograms are aligned in RELION4 and the 2D subtilts are aligned in RELION5
Below is my two extraction commands, in case I forgot a 2D specific parameter or similar. 2D:
WarpTools ts_export_particles --2d --settings warp_tiltseries.settings --input_star ${input_starfile} --coords_angpix 1.7005 --output_star relion/particles.star --output_angpix 6.802 --box 64 --diameter 250 --perdevice 23D:WarpTools ts_export_particles --3d --settings warp_tiltseries.settings --input_star ${input_starfile} --coords_angpix 1.7005 --output_star relion3d/particles.star --output_angpix 6.802 --box 64 --diameter 250 --perdevice 2For context: I'm converting a larger dataset from Warp 1.0.7 to WarpTools and everything to the point of subtomogram extraction has worked perfectly. I imported averaged frames, redid ctf estimation, imported mask files from old boxnet, imported old tilt alignments, redid 3D CTF (+ tomogram extraction for reference), and everything looks perfectly well.