Patch motion correction for .eer files

Hi, all. I met a problem about patch motion correction. I collected my dataset with FEI falcon 4i and got .eer files.
And the result of motion correction may be not normal. The alignment of raw coarse and smooth coarse was good. However, the smooth fine curve is very different with the others.

Btw, parameters of import movie job are there:

Can anyone help me solve the problem?

What happens if you set the Fourier cropping to 1 instead of 0.5, and the EER upsampling to 1 instead of 2?

You are right! Thank you very much. I got it.

Great! Good luck with processing!


I have similar issue, I tried the suggestion above, did n’t help. Kindly advise.


This is on krios.

Also with EER data? What settings did you use?

Yes these are EER data from Krios

This is for Patch Motion Correction

I got this very blurry initial 2Ds

I have screened this previously using Glacios, had no issues. Please see parameters for the screening attached.

Same grid used for both datasets? Or different grids?

With so few particles it is difficult to conclude whether or not the dataset is poor; early iterations and low particle counts usually look pretty bad. Although that oscillating waveform in the penultimate class worries me a little.

These were different grids. This picture is one odd one with less number of particles, but there are a lot of particles in other grids. The abinitio volume looks correct but is 3 times smaller than the abinitio volume from Glacios. Upon increasing the size the model fits the map correctly. What could be the reason for this?

Ab initio volume being so far off indicates something wrong with the settings used for import, but nothing stands out as wrong in the screenshots above. You’ve double checked the pixel size is correct (for 4K rendering)?

What does the 2D look like when particle count is higher? Can you see distinct features or are they still messy low resolution blobs?

  1. [quote=“rbs_sci, post:12, topic:9837”]
    You’ve double checked the pixel size is correct (for 4K rendering)

Could you please elaborate on this?

  1. I changed the Patch Motion Corr. output F-crop to 1/2 for the Krios data - It looks like that from Glacios. The abinitio volume voxels increased to 320 and for the Glacios one it is 300.

BTW both data are short collections (400 img) from Glacios and Krios.
I have not collected full data for the Krios one yet.
Any clue on why this would be happening?
Thank you!

EER has three possible rendering modes - 4K (upsampling 1), 8K (upsampling 2) and 16K (upsampling 4). With RELION, you need to manually adjust the input pixel size based on what EER upsampling you choose, while with CryoSPARC you always want to give it the upsampling 1 pixel size, and it will re-calculate from there.

For picking/sanity checking particles, those boxes are way too big for the “particles” you are selecting. Also, none of those classes look like actual particles. Can you provide an image of what you’re picking from the micrographs?

Thank you for the advice regarding EER.

Extracting particles at box size 360 bin size 360 (Taken from the manual picker log)
What size do you suggest, (particle max diameter 125A as observed in the structural model)

2D select

What are your recommendations on 2D and abinito setting?
thanks a lot!

For templates for autopicking those classes are usable. Try template picking, see what it picks. Ab initio I usually leave as default unless working on something high symmetry to prevent pancaking. What I change for 2D depends, but I usually now default to the following:

Classes: [varies; usually 100-200 depending on dataset size]
Initial classification uncertainty factor: 5 [default 2]
Number of final full iterations: 2 [default 1]
Number of online-EM iterations: [varies; try to cover dataset at least 150%]

If the dataset is extremely heterogeneous, I’ll increase Batchsize per class, but this could be anything from 200-2000.