Preparing CryoSPARC-Processed Files for Futher Processing in Relion?

Hi all,

I have been quite happy with my high quality map results from doing all preprocessing (motion correction, CTF estimation) in CryoSPARC, but I now want to pull my data into Relion just to compare and see if the map resolution and quality is enhanced further. I’ve been told I need my processed micrographs, a micrograph .star file, a file of my final particle set converted from CryoSPARC .cs file, and my CryoSPARC map for an initial Relion model. I’m a little confused about how to best go about this, so here are my questions:

  • For the .mrc files, do I just use the output micrographs from my Patch motion correction job? Is it alright that I did Patch CTF estimation downstream of this; essentially, is this where my processed micrographs are stored?

  • For the converted particle file, should I use the output .cs file from my 2D Select job or from a refinement? I have heard that it can be better to use an output particles.cs file from a refinement but I’m not sure why. I also get a error when I use an exported refinement output file, but not when I use the exported 2D select output file.

  • How do I get a file when I did all of my preprocessing in CryoSPARC? Can I just convert the micrographs.cs file within the exported Patch Motion Correction Job using I tired this and got failed results, so i’m wondering which exact file and commands to use to properly do this, or how else I can get a file.

If I’m super off-base and there’s a better technique to getting further refinement in Relion going, please let me know what I need to do and/or what different set of files I need to bring into Relion.

Thank you!

Hi amcc,

Depending on what types of Relion jobs you want to run, you have a few options. If you would just like to try and refine the Euler angles and transnational alignments, I believe there is a way to simply export the particle stack as an .mrcs (the particle “blob” file is actually a mrc stack) and using pyem you can carry over angular priors for refinement in Relion if you use the cs file from your 3D refinement. The advantage of using the 3D Refinement .cs file is that it will have Euler angles associated with each particle that correspond to their alignment to your refined map, while a 2D classification .cs file will not contain this information.

However in my opinion, if you are moving to Relion you should try Bayesian particle polishing as this often produces better motion correction results than patch motion. Others may have a more efficient workflow, but I re-run motion correction and CTF estimation inside Relion because the motion correction trajectories cannot be not exported from CryoSPARC by pyem. Then you can generate a star file from your 2D classification or 3D refinement in CryoSPARC using both the paticles.cs and passthrough-particles.cs (if present) . The star file _rlnMicrographName column will need to be edited to point to the CtfFind micrograph directory instead of the PX/JYYY/ cryosparc directory. You can “re-extract refined particles” using the that was generated in Relion and the particles star file that you generated with pyem. This will allow you to run any Relion job on your curated particles from CryoSPARC.



It’s better not to get ahead of yourself with complex processing - wait until your resolution seems limited by CTF to do CTF refinement, wait until it’s limited by local motion to do polishing, wait until it’s limited near-Nyquist to unbin. It can be hard to stay disciplined in this way, for me too, but it will give you faster results, and you’ll be more confident about what needs to be done next.

Here’s the how-to on setting up your new Relion project based on cryoSPARC results.


@Ablakely how did you do to edit the _rlnMicrographName?

Thanks in advance

Hi Fernandes,
I used expression matching in the VI text editor.


This will perform a search and replace on every line of the star file. You can use a back slash as an escape character if you need to include forward slashes in string1 or string2.

1 Like