Downsampled particles csparc2star

Troubleshooting
1

I am trying to make a star file from downsampled particles (import, downsample, homo refine), but the downsampled image names in the resulting star file do not correlate to the original names, even when I include the downsampled_particles.cs and downsample passthrough.cs files in the csparc2star.py command.
I ran the command as: csparc2star.py homoref.cs homoref_passthrough.cs downsample.cs downsample_passthrough.cs newfile.star
The output newfile.star has imagename as J1/downsample/batch_00000_downsample.mrc
Using the particle file from the job export folder did not work.
Any way to get this to work? I only downsampled (to the same box size) to prevent csparc from importing the whole particle stack to ssd cache.

2

Run a homogeneous reconstruction job, using the outputs from your refinement, but using the low level results interface to swap the particle input from the downsampled particles for the original. Then take the resulting particle.cs file and try converting that?

Cheers
Oli

1 Like
3

Wonderful. This worked. I just had to replace the input particles.blob with that from the imported particles and csparc2star gives original imagename. Thanks!

4

Another issue I’ve run into is that I want to crop the particles with relion after downsample in cryosparc. I need to keep the original image name so I can unbin after 3D classification, but need the new “downsample” name present to run relion_preprocess.
If the csparc output from homo reconstruct keeps a line for imageOriginalName then no problem and I can revert back. Not sure if this will be the case as it’s still transferring the 300 gb of unbinned particles to cache to run the reconstruct job with imported particles.blob.

5

The resulting star file does not include the new downsample particle stack names, so it seems downsampled particles cannot be cropped.

6

@user123

  1. You can use csparc2star.py --boxsize to adjust aligned particle shifts if the refinement box is not the same as the particle box. I believe the downsample job takes care of it automatically, but I think it’s needed if you manually set a refinement to have a smaller box than the particles.

  2. If you are having trouble getting the image names and alignment parameters you want together I would suggest you explore star.py, also in pyem. The --copy-X options and --merge-source <file> --merge-key <key1,key2,...> --merge-fields <field1,field2,...> options may be of use. There is also nothing wrong with using a text editor to search replace the paths on the image names.

7

Hi Daniel,
Thanks so much. I’ll explore option 2. I was not aware of the merge source option. For now, I re-extracted the particles, but this will be helpful when mics are not readily available and I need to go between polished particles and binned particles.