Using particles.blob accurately

Local Refinement
1

I am currently running a local refinement job using 2x binned particles, which were preceded by symmetry expansion and follow up 3D classification. I combined the best 2 classes for the local refinement. I want to run a followup local refinement using the same particle stack but unbinned, and I have run an extraction job with no binning. My questions is - how do I run this local refinement without having to do 3D classification again? I could symmetry expand the unbinned particles, but I only want a subset of them to feed into my final local refinement (corresponding to the best classes from my 3D classification on the binned particles). Would using the particles.blob output from my 2nd symmetry expansion (unbinned particles) and adding that that to my new local refinement (using the first local refinement particle stack as input) work to achieve this? Thanks in advance!

Flow

2

Hi @flow_proz241! Could you confirm that this is your desired workflow?

  1. Start with some particle stack
  2. Perform symmetry expansion on this stack
  3. Classify, local refine, etc.
  4. “unbin” the particles from step 3, the alignments you found in the local refinement step
  5. Perform a Local Refinement on the full-size particles from step 4
3

Yes, this is it! I have found that when I use the particles.blob from the unbinned particles that I symmetry expanded, it undoes the symmetry expansion, and I haven’t found a workaround for this yet.

4

Hi @flow_proz241, sorry for the delay! You’ll have to use cryosparc-tools to do this – here’s an example (you’ll also need to install polars for this script). Running the script like so:

./reverse-downsample.py http://localhost:39000 P392 J483 J480

where

  • Your CryoSPARC server is at http://localhost:39000
  • P392 is your project UID
  • J483 is the job you want to keep particle metadata from (in your case, the Local Refinement)
  • J480 is the downsample particles job

will create a new External Job with the particles from J483, except with full size images.

5

Would this work if I never extracted the particles at the full box size? All of my workflow so far has been on particles extracted with fourier crop x2, I didn’t extract at full box size and downsample later.

6

Ah, in that case the easiest thing to do would be:

  1. Extract the original (not-expanded) particles at full size. Ensure that Recenter using aligned shifts is turned off so that the particle images match the expanded particles
  2. Run any job with them (perhaps a Homogeneous Reconstruction Only)
  3. Use the job from step 2 here where the script asks for a Downsample Particles job.

However, unless you have a very high symmetry order or very constrained storage space, it may be best to simply extract your expanded particles instead.