Phase shifted 2D classes?

I collected a dataset of several images and picked particles from these micrographs (that all visually look normal) but I got 2D classes with both black and white averages. When I looked at the curate exposures, it seems like the black averages are in the majority from phase shifted micrographs. There were no phase plates used. Does anybody have any idea what is happening? Thanks, Chiara

Hi @aglape,

Thanks for posting. We also saw the images you posted on twitter (presumably from this same dataset). As was pointed out there, most likely the issue is that phase shift estimation was turned on during CTF estimation, but this can cause issues if the data was not actually collected with a phase plate. When there is no phase plate present, the phase shifts should be zero for all images, whereas the plot you posted shows that some micrographs had ~0 and some ~180 degree phase shifts.
If you re-run CTF estimation with phase shift estimation off, this should fix the issue. Let us know what happens.


Hi Ali, how do I do that on cryosparc? How do you turn off the phase shift estimation? I didn’t turn it on in the first place. I assume it’s the phase shift refine. The same thing happened when I ran CTFfind…
Any idea how to fix it? Can I invert the contrast of the particles to try and use them in the main job? I don’t see an option to do that.
Any help would be greatly appreciated!

Hi Chiara,
Both cryoSPARC and CTFFIND only output non-zero phase shifts (like the X-axis of the plot from your tweet) when optimizing phase shift is explicitly turned on the by the user. So where did those values come from? In cryoSPARC it’s one of the sliders in the Patch CTF job, should be off by default.

Hi it was off. I don’t know where this shift comes from…

Hi @aglape,

Just to confirm, in CryoSPARC, when you run Patch CTF, the phase shift will be optimized (i.e. it will not remain zero) if either:

  • At import time, you turned on the “Phase Plate Data” switch in the import job. This setting is carried forward with the data throughout processing
  • In the Patch CTF estimation job, you turn on the “Do phase refine only” parameter. This forces the job to refine the phase shift (whether running on phase plate data or not)

Otherwise, the phase shifts of all micrographs (and particles) should be zero. If you didn’t set either of these parameters, it’s definitely very strange that the phase shifts are non-zero. Especially so if the same thing happens in both Patch CTF and CTFFIND.
As a quick test, can you try to re-import just a few (perhaps 20) of the movies into CryoSPARC and re-run motion correction and CTF estimation on this small subset with all default parameters, and see if you still see non-zero phase shifts after that?