I used cs2star to convert particle stacks from several NU-refinement jobs and imported them directly into RELION for refinement and 2D classification (without re-extraction). However, after refinement, the output MRC file (over 200 MB) appears empty, with no visible density.
To verify whether the particle stack is correct, I reconstructed it using the same particle.star file with the following command:
mpirun -n 24 `which relion_reconstruct_mpi` --i particles.star --o map.mrc --ctf --skip_gridding
This produced a good map. However, when running 2D classification, the resulting class averages contain only noise (see attached image).
Does anyone know what might be causing this issue? Any suggestions would be greatly appreciated.
What options did you use in csparc2star? Did you use –inverty and –flipy? I think that if you use the extracted particles from cryosparc directly, then you do not want to use those parameters in csparc2star. Inverty flips the y-coordinate, that won’t matter unless you reextract in relion from MotionCor2 micrographs but flipy will flip the sign on the defocus angle. I would think that should also mess up the reconstruct job you ran but let us know exactly what you did.
For cs2star, my command was “cs2star /path_to_cryosparc/JXXX -p”. Initially, I suspected that the cs2star job might not have completed properly, so I also tried csparc2star using “csparc2star.py path_to_cryosparc/JXXX/JXXX_011_particles.cs path_to_cryosparc/JXXX/JXXX_passthrough_particles.cs csparc_JXXX_particles.star”. I did not use the “–inverty” or “–flipy” options.
However, both methods produced the same issue. In each case, I also generated the corresponding particle stack using “relion_stack_create –i particles.star –o particles”. Interestingly, I can reconstruct a correct map directly from both .star files using relion_reconstruct, which suggests that the particle stack itself should be valid.
The problem arises during downstream processing — 2D classification produces only noisy class averages (as shown in my first comment), and 3D refinement results in a map with no visible density.
Does anyone have any insights into what might be causing this issue? Thanks very much!
Ok we need more details about what you are doing in relion then.
Next thing to do is to check the star file. Does it have all the columns that you would expect. Defocus values, three angles, and two shifts?
Are you running Class3D without alignment or with alignment? What are your settings?
How is Class2D being run? I have seen some cases where 2Ds look like this with bad settings and look fine in cryosparc for instance.
I figured it out! It turns out that my imported particles didn’t work properly in RELION 4 but worked fine in RELION 5. After running the process in RELION 5, the results look normal now!
Thank you so much — your suggestion really helped me realize it might be a version issue.
1 Like
