I/O error during Homogeneous Refinement after re-extracting unbinned particles

3D Refinement
1

Hi ,I encountered an I/O error when running Homogeneous Refinement after selecting particles from 2D classification and re-extracting them at the original pixel size.

My workflow was:

Motion Correction → Patch CTF Estimation → particle filtering → Blob Picker → particle extraction with 2× Fourier downsampling → 2D Classification → select good 2D classes → re-extract the selected particles at the original pixel size → Homogeneous Reconstruction only.

For the final re-extraction, I used:

  • Extraction box size: 384 pixels

  • Fourier crop to box size: left blank (no Fourier cropping)

  • The purpose was to recover the particles at the original pixel size.

The Homogeneous Refinement fails with the following error:

image
image1120×650 9.71 KB

The 2D classification itself completed successfully. The error only appears after selecting the 2D classes, re-extracting the particles without Fourier cropping, and using those particles for 3D refinement.

Could this be caused by a mismatch between the particle .cs metadata and the original micrograph/movie metadata?

In particular, I would like to know whether the problem could be related to:

  1. The selected particles retaining incorrect movie/frame metadata after 2D classification;

  2. The re-extraction from the original micrographs;

  3. EER fraction metadata;

  4. Particle cache or MRC files;

  5. A mismatch between the input particles and input micrographs/exposures.

Is there a recommended way to diagnose which particle/exposure is causing this problem?

Thank you.

2

@chaoyang May I ask:

  1. What version of CryoSPARC do you use.
  2. Have you tried, as a diagnostic step, to run homogeneous refinement with the Cache particle images on SSD parameter disabled?
  3. What is the output of the command
    df -hT /home/u18601538/software/cryosparc/cache
  4. Is the filesystem where /home/u18601538/software/cryosparc/cache significantly faster than the filesystem where CryoSPARC projects are stored?

As an aside: Because caching tends to entail the frequent writing of large amounts of data, the wear on underlying storage devices can be significant. Device wear may lead to failure of the storage, which, in case of cached data, is a minor problem because a primary copy of the data may still exist (in the project directory), unless the primary copy stored on the same device as the cache. For this reason, one may want to store the cache on separate devices from more valuable data, like raw movies, CryoSPARC project directories, or the CryoSPARC database.

Gentle reminder: Please post error messages as text instead of as a screenshot to enable users with a similar problem to more easily find this discussion.