Bug Report: multiple masks imported, only one passed through to new jobs

Import and Export
1

Hi CryoSPARC team,

I continue to experiment with the viability of using CryoSPARC for working with very large subjects, despite the limitations imposed by PyFFTW on box size. Stumbled onto another issue…

I took advantage of the ability to import multiple volumes at the same time with the wildcard method but something very odd is happening…

…

All three volumes are masks. They appear to import correctly (imported_mask_1, imported_mask_2 and imported_mask_3) and all look correct and distinct across X/Y/Z slices. When creating local refinements, I pass mask_1 to the first block/particle set, mask_2 to the second block/particle set and mask_3 to the third. Graphically, all appears OK. Entries indicate that mask_1, mask_2 and mask_3 are connected and loading appropriately and the little visualisation correlates to each independent mask.

If I download them again and check them individually they are the correct mask.

The problem comes when the refinement starts:

For all three local refinements, mask_3 is loaded and used.

I’ve tried both drag-n-drop of the “normal” output tiles, and the “dig through the Outputs, directly link volume.blob” route.

If I load each mask in a separate job, everything works as expected.

Cheers,
R

…

For other CryoSPARC users:

If anyone is using the wildcard method to load lots and lots of maps or masks simultaneously, please check that CryoSPARC is actually correctly loading the one you think it is!

1 Like
2

@rbs_sci Thanks for your post. May I ask:

  1. Which plot or log entry, at which iteration, indicated that mask_3 was used in all cases?
  2. What custom parameters did you set for the local refinement jobs?
3

As soon as the Local Refine starts, the mask loaded is the wrong one, and masked half-map A and B have the wrong mask applied.

Custom params:
Use pose/shoft gaussian: ON
Std. Dev. Deg: 2.5
Std. Dev. Ang: 2.5
Max alignment res. (deg): 0.05
Recenter rotatations: ON
Recenter shifts: ON
No. extra final passes: 3

But the same happens regardless of settings used.

4

Thanks. Please can you confirm that I understand correctly:

  1. The Input Mask Reals Space Slices plot, below the Noise Model Initialization (1/2), corresponds to the expected mask.
  2. The Real Space Mask Slices Iteration 000 plot corresponds to the wrong mask in two out of three cases?
5

Not quite.

When importing masks, everything looks OK.

  1. When running Local Refine, right from the first step the mask loaded is incorrect (and the first of the imported masks).

  2. Correct.

6

Thanks. Please bear with me (and my questions) as I am trying to reproduce the issue.

Please can you post a screen shot that clarifies:

  • whether you refer to the entries and visualizations of the Volume Import or Local Refinement job
  • where in the UI and under which tab those “correlating” visualizations appear?

Would you be willing to share with the developers:

  1. the job report for the Import Volumes job
  2. job reports for the three Local Refinement jobs
  3. output of the command
    find /path/to/project/JX -type f -exec shasum {} \;
  4. *cs files inside /path/to/project/JX

where JX is the Import Volumes job ID. I will send you a direct message with the email address where you could send this information.

1 Like