Reference motion correction

Hello, we would like to do reference motion correction in v4.4 and have several questions:

We did motioncorr2 in relion, import the corrected micrograph in cryosparc and extract particles, do signal subtraction and get the local refined map and particles. As discussed previously, the patch motion correction in cryosparc with imported TIF movie flip the y axis. Therefore, we have re-extract particle with alignment informations by using ‘flip mic y’ option to resolve this flip. I was wondering:

  1. Will this flip corrected automatically in the new reference based motion correction since when I run the inspect particle job, the flip is not considered.

  2. If answer to 1 is NO, what should we do to resolve this? For instance, for location/frac_x and location/frac_y in cs file, how should we change the coordinates to reflect this?

  3. Should the corresponding ctf, alignments3D changed accordingly in some parameters? For instance, phase-shift angle in ctf, shift in alignments3D, etc?

Thanks a lot.

Hi @nym2834610,

Thanks for your interest in using reference-based motion correction (RBMC) in CS v4.4! The current implementation of RBMC does not support outputs from MotCorr2. RBMC requires background estimates (low-level output called background_blob) and MotCorr2 does not provide these as outputs (even if you were to run MotCorr2 from within CS). To perform RBMC, you will need to perform the following steps:

  1. Import raw movies
  2. Patch (or Full Frame) motion correction, as implemented in CS
  3. Re-extract particles with known locations from previous processing
  4. Ab-initio or Homogeneous reconstruction
  5. Homogeneous or NU- refinement
  6. RBMC
1 Like

We did run patch motion correction in CS. What we noticed is when importing TIF and perform patch motion correction, the movies and motion corrected mrc files are y-filpped compared to the Motioncorr2 job performed in RELION previously. Because our data processing requires multiple signal subtraction step, we have to get alignment information from the previously processed .cs particle metadata file.

To resolve this, we first try applying the ‘flip-y’ option in extract-partilce job in CS and could get the particle images and get similar reconstruction results. However, this y-flip could not be done in RBMC. Therefore we are currently trying to do ‘flip-y’ in the TIF movies, patch-motioncorr and patch ctf and then try to get reconstruction for RBMC. If possible, brief meeting will be very helpful to get some help.

Hello, we have successfully flipped the TIF images by imod command:

clip flipx -2d X.tif X.mrc
mrc2tif -s -c 5 -P X.mrc X_yfliped.tif

Taking the metadata cs file from cryosaprc refinement jobs (extracted from the motion corrected micrograph imported to cryosparc from relion), copy the alignments3D, ctf, location, filament, etc. prefixes using cryosparc_tool python module and re-extract the particles in cryosparc could couple the previous alignment results with cryosparc patch-motion-corr job for the later RBMC process. The overall resolution is almost identical.

This will produce equivalent particle images compared to relion-based-motioncorr2 micrographs when taking the previously processed cs metadata file from cryosparc that based on importing