I started my data analysis on .mrcs preprocessed by the EM center’s automated pipeline, and I now want to try RBMC to push resolution a bit more before local refinements of especially troublesome areas. However, it seems that the raw movies are inverted in Y from the preprocessed .mrcs: after reassigning particle locations to the motion-corrected and CTF-estimated .mrcs from the raw movies, I was able to extract my real particles only when flipping in Y before extraction. This has not helped with RBMC, however – even though I use the exposures output by the “flip mics in Y before extract” extraction job, RBMC is still stubbornly trying to track the motion of noise instead of my particles.
Is there an option to flip my freshly processed movies in Y in the cryoSPARC interface? If not, is there any possible workaround for this issue?
As you pointed out, Y-flip during extraction won’t update particle coordinates. For flipping the locations in Y, I’ve used csparc2star.py.
Run csparc2star.py on .cs files from a refinement job of your choice. The recent versions of csparc2star.py will apply the Y flip by default during .cs → .star conversion.
Import the .star particle stack to Cryosparc. Make sure to connect the source micrographs as input for the import job to keep the flipped locations along the particle stack.
Check with Manual Picker to confirm you have the correct locations.
Extract the particles with the updated locations with Force re-extract CTFs from micrographs: True.
Optionally, you can double-check the re-extracted particles by running them through a refinement / homo. reconstruction before running RBMC.
The initial refinement after reassigning to the source movies lost ~0.3A resolution compared to the final refinement from the preprocessed .mrcs, HOWEVER, the refinement post-RBMC more than made up for it – went from 4.64A to 4.32A!
Will apply this procedure to my other top refinements – thanks again kookjookeem!