Cryosparc to relion: Few particles in groups

Troubleshooting
1

Hello Colleagues,
We have exported particles and a volume from cryosparc homogeneous refinement to relion for 3D classification. With the first dataset, there were no issues. However, for our second dataset, relion is crawling in speed in its 3D classification. Relion is complaining that there are very few particles in each group and that I should combine them together. I have two questions-

  1. How are the number of particles per group assigned/controlled and would it be different for two datasets?
  2. How should I fix this? Thank you.
2

Ok, I found the answers to the two questions:
https://relion.readthedocs.io/en/release-3.1/SPA_tutorial/Class2D.html
https://www3.mrc-lmb.cam.ac.uk/relion/index.php/Grouping_procedure
It looks like every micrograph is a group. I have ~18,000 micrographs and ~1,000,000 particles. So that is ~55 particles/micrograph. Yet, RELION complains that there are only 2-4 particles per group. How can I avoid this issue? Can someone please comment on this? Thank you.

3

After an initial round of refinement in relion, you can regroup particles into less groups using the subset selection tool (with the model.star file as input). The resulting particle.star can then be used for a fresh refinement.

4

Thanks Oli. Yes, I read that suggestion one other place as well. The purpose of all of this is to take the particles and volume from my cryosparc/homogenous refinement job into relion/3D classification (without alignment). As of now, relion/3D refinement is crawling slowly with this dataset (particle boxsize=80x80A, 3.8A/pixel) and is taking hours. Hence, I was looking for alternative ways to avoid this issue of too many groups from the very start. Please do let me know if you have any suggestions.