I combined two data sets with slightly different pixel sizes (1.083 vs 1.073 Å). The latter group was collected using 45º and 30º tilt. Using MotionCor2 (through Relion) I adjusted the pixel size of the the 1.073 Å micrographs using F-box/Fbit ( input value = 0.9907663896583560) to try and make all the micrographs at pixel size 1.083 Å. Micrographs were then imported into CryoSparc for Patch CTF, template picking, and particle extraction with downsampling (extraction box size: 440, cropped box size: 220).
The correction seemed to work well enough for CryoSparc to merge the datasets for 2D classification without throwing any errors about pixel size differences. I went through some preliminary 3D classification with the downsampled particle stacks and everything looked OK until I re-extracted particles without downsampling (extraction box size: 440). The NU-refinement on those particles gave me this FSC curve with a extra bump.
From reading some posts, it seems like this can happen for some transmembrane proteins, but my particles are ribosomes and I’ve never seen this happen before.
These posts ( Double peak on FSC Plot - #7 by dendena + Bumpy FSC curves ) suggested that this extra bump may be caused by the wrong pixel size or beam tilt. I’d love input on what others think the issue is and how to resolve it.
Combining particles in this way requires the pixel sizes to be very precisely known. It may be best to process the datasets separately, refine magnification/Cs, and use the refined magnification in resampling, instead of the nominal value.
Hello! I found that “total_dose_e_per_A2” was off by a decimal place for one dataset. It was 5.229 but should have been 52.29. Could this explain the issue?
Beam tilt is the usual culprit. Also there looks like there are some duplicates hiding as FSC doesn’t reach zero. Dips in the FSC from detergent/micelles usually don’t manifest at higher frequencies - usually it’s quite broad around 6-8Ang…
Hello! Could you elaborate on this please? I feel this might be similar to another suggestion I’ve received, which was to combine particles in Relion for 3D classification after initially processing them separately.
