Cannot replicate refinement from cryosparc after csparc2star

Troubleshooting
1

I am encountering a strange issue after running csparc2star. My goal is to extract particles without normalization after running standard processing in cryosparc. Since cryosparc cannot do this, I must extract in relion. I am not extracting without normalization, yet, just testing the pipeline with normal settings that should be consistent with cryosparc processing.

My current issue is that when I run csparc2star.py, extract particles in relion, then import those particles to cryosparc and run a similar refinement, it is 1 Å worse than the original refinement using particles from cryosparc. This seems to not be a problem with relion’s particles, because when I reassign the particles to the original micrographs and extract with cryosparc, the refinement is also 1 Å worse.
The original cryosparc refinement did not have higher order ctf optimization. I know csparc2star does not carry this over, but that’s not the issue.

I expect that particles from cryosparc should go to relion and then back without such a drastic resolution drop. Here’s my pathway:

  • Import tiffs, Cryosparc patch motion & ctf, particle pick, extract, 2D, extract unbinned, refinement direct from extract with optimize defocus only, goes to 2.03 Å

  • Run motioncor2 in relion, import mrc to cryosparc

  • Reassign particles from 2.03 Å refinement to new micrographs from motioncor2

  • Extract with binning, Homogeneous refine to re-center

  • Extract without binning, refines to 2.04 Å
    (no issue here, this shows the same particle coordinates extracted from cryosparc patch motion and motioncor2 both refine to 2 Å with pure cryosparc processing)

  • Csparc2star.py & extract in relion from the same motinocor2 mics

  • Import particles to cryosparc

  • 2D classification test, particles look nice

  • Homogeneous refinement, same settings optimize defocus only, Refines to 3 Å

  • Reassign particles to motioncor2 mics, extract, refines to 3.0 Å with same settings

Another test indicating a problem - when I run homogeneous reconstruction on the imported particles from relion, they go to 8 Å, so something is very wrong.
When I import the starfile from csparc2star from the original cryosparc refinement pointing to the cs-extracted particles from patch motion or motioncor2, it reconstructs the expected resolution 2.08 Å, so csparc2star maintained the alignments/shifts of that original refinement and cryosparc reconstructs them just fine after passing through csparc2star.py.
But those relion extracted particles reconstruct & refine to poor resolution.
When I run reassign particles from this import particle job to the motioncor2 micrographs, extract and refine, it goes to 3 Å

2

It seems like one issue has to do with the requirement for flipy when comparing particles from cryosparc and relion. Flipy is not required to reconstruct a csparc2star file that points to the original cs-extracted particles, but is required to reconstruct a csparc2star file that is used to downstream extract particles in relion.

This still doesn’t explain why refinements of any relion-extracted particle gives 1 Å worse than the cs-extracted particle for the same set of coordinates.

The shape of the 3.0 Å fsc curves is a sharp drop and maybe suggests the need for beamtilt refinement or other higher-order ctf parameter optimization, but this was never performed for the J806 refinement, which only had defocus optimization.

Data collection was fairly standard krios with K3 at 0.83 Å/px and Cs 2.7.
All extractions are performed on the same set of micrographs from motioncor2

cs_problem_equineapof
cs_problem_equineapof1920×1197 115 KB

3

Have you read this thread (Experimental support for Relion's Bayesian polishing in csparc2star.py - #67 by olibclarke)?

I wonder if it could be due to the orientation of the relion-extracted particles not matching the defocus angle - so effectively the astigmatism is wrong?

In reference to this:

  • Reassign particles to motioncor2 mics, extract, refines to 3.0 Å with same settings

Was this using the particle CTF estimates (presumably derived from Patch CTF on Patch Motion mics) or did you rerun Patch CTF on the mc2 mics? If not, that might be worth doing (and then make sure to extract using the micrograph CTF estimates) - if the issue goes away after running Patch CTF on the mc2 mics, that would be informative.

4

Hi Oli,
You’re right, I finally got the relion extracted particles to similar resolution by running patch-ctf on the motioncor2-mics and patch ctf extract.
When I import the motioncor2 micrographs, reassign the old particles extracted from CS-patch-motion, and extract in CS with Flip-Y, the refinement goes to 2.04 Å, so it seems cryosparc does this transposition internally.
Thanks!!

1 Like