Strange results after local motion correction but patch motion correction works fine?

[iga_k]

Hello everyone,

I am processing a dataset obtained using Titan Krios with Falcon 4i camera and Selectris energy filter. It is a brand new microscope.

I am processing using cryosparc 4.4.1.

I am running patch-based motion correction and patch CTF estimation.

Here are my 2D classes:

I obtained 2.8A map with NU refinement which shows all the expected features for this resolution.

I am trying to improve resolution slightly by running local motion correction on my final set of particles. Job finishes without an error.

However, the obtained set of particles shows some strange artifacts.

2D classes after local motion correction:

Attempting refinement on the particles from local motion correction job results in a map that is unusable:

Since then, I processed different datasets from the same microscope and they all show the same effect (patch motion - good, local motion correction - strange).

We also have Glacios with Falcon 4i camera and Selectris energy filter. Collecting datasets on Glacios from the same set of grids and processing with cryosparc/patch-based motion correction and patch CTF estimation leads to a reconstruction at 3A. Local motion correction further improves this dataset to 2.9 A.

It seems that the issue is with datasets from our Titan Krios with Falcon 4i camera and Selectris energy filter.

If anyone has any suggestions on how we can make local motion correction work with our setup, your input would be greatly appreciated. I have been trying to adjust the settings for local motion correction but have been unable to improve the results.

[rbs_sci]

Which of those classes did you take forward for NU refinement? There are several there (mostly later ones) I would dump on principle…

When you say “local motion” do you mean the (now effectively legacy) Local motion correction job, or the new “Referenced Based Motion Correction” job?

I’ve never had much luck with the former, but the latter usually yields good results.

Other than that, the ringing and lines in some of the second round of 2D classes looks like bad defect correction. Check your gain reference, check your defect maps (the Falcon camera tool will allow you to check the defect list) and check what versions of CryoSPARC and EPU you’re running. Earlier releases of CS with RBMC had a bug regarding defect files (fixed in later releases) and the gain reference that EPU includes by default in some of the earlier Falcon 4 supporting releases would not work properly.

I’ve never seen this from our Falcon 4(i) and Selectris-X, but I have seen it from K3 data with several uncorrected defects.