Relion 5 tomography polishing generates a miniaturised map inside the main map

[mathewmclaren]

This is a template for reporting bugs. Please fill in as much information as you can.

Describe your problem

I've been running the Relion 5 tomography pipeline using a dataset very similar to EMPIAR-11888, an ordered s-layer assembled onto exosomes. Processing this dataset using Warp > Relion > M achieved a resolution of around 11A with around 7800 particles (C3 symmetry). I've now rerun all the steps using only Relion 5 and currently have a map with a similar resolution (also C3 symmetry) and wanted to try polishing. We find that the steps in M are usually effective around this resolution so assumed that polishing/tilt series refinement would give a small improvement.

No per-particle motion was run and align by shift only was set to no (full command is below) then I ran a reconstruction with exactly the same parameters as prior to polishing. The resolution dropped slightly, but most importantly I noticed an artefact in the map. It appears as though there is a smaller version of the s-layer in the middle of the pore. The below image shows the the original map (right, yellow) and the new map (left, grey) in ChimeraX. It is clear that the left map now shows the same features in the middle that exist in the main map.

I roughly measured the diameter across the hexagonal subunits and got a value of ~60A, compared to ~180A for the "full" map. I assume this is in some way related to the 3x binning used for the particle extraction pre and post-polishing. I checked the output star files from the Polish job and couldn't see anything obviously wrong. I can post these if necessary.

Environment:

• OS: Centos 7
• MPI runtime: OpenMPI 1.10.7
• RELION version RELION-5.0-beta-2-commit-a0b145
• Memory: 256GB
• GPU: Nvidia RTX 3090

Dataset:

• Box size: full - 128px / cropped - 96px
• Pixel size: 1.63Apx (binned to 4.89Apx)
• Number of particles: 17,000
• Description: Assembled S-layer protein

Job options:

• Type of job: Polish/Reconstruct
• Number of MPI processes: 5
• Number of threads: 8
• Full command (see note.txt in the job directory):

`which relion_tomo_align_mpi` --i Refine3D/job108/run_optimisation_set.star --ref1 Reconstruct/job109/half1.mrc --ref2 Reconstruct/job109/half2.mrc --mask MaskCreate/job111/mask.mrc --fsc PostProcess/job112/postprocess.star --theme classic --o Polish/job114/ --b 128 --r 5 --only_do_unfinished  --j 8  --pipeline_control Polish/job114/

[mathewmclaren]

Just a quick update on this. I reconstructed my map pre-polishing at bin1 and reran the polishing. I then reconstructed at bin3 then postprocessed and saw that the resolution improved from 11A to 10A. I also don't see the mini version of the map in the middle any more. There must be some issue regarding the binning of the reconstruction map used for polishing, but I find it odd that the map shows a combination of full size and scaled features in the same map.

[scheres]

It is essential to use a bin1 reconstruction for running CTF refinement or polishing.