My apologies, by “connect motion-corrected micrographs instead of movies”, I meant in the cryosparc-tools script, not the reference motion job! From the rest of your message, I can see you did not do this.
In your Python script, rather than specifying the “Import Movies” job UID for old_import_movies_job_uid, specify the UID of the Patch Motion or Patch CTF job that you connected to your Reference Motion job.
Replace the “imported_movies” output name with either “micrographs” or “exposures”, the output name of the parent job.
Replace both lines that begin with slots= with this:
I now understand your suggestion. As you outlined, I have tried using the “patch motion correction job” as the input for the external job. In short, it worked-- the “reference based motion correction” can now run, at least for computing the cross-validation scores. I will report later whether the job can successfully finish.
FYI, I had additional issues with this approach. Namely that new movies are not complete and the initial “patch motion correction” had failed jobs. Thus, there are exposures in the new folder but not in the “patch motion correction” and vice versa. Using python, it seems possible to locate the common files and limit the output to these files. cryosparc-tools is great!
The “reference based motion correction” job is now doing the final local motion correction.
However, the plots for “FCC model fit” and “empirical dose weight” appear blank (below). Is this a bug for Falcon4 eer file? Are there ways to check whether the “FCC model fit” is in place?
@nfrasser, the job finished. But the reconstructed particles are wanky. See the display of particles with RELION below. Do you know what was the cause and how can we fix it?
Hi @sunch, the problem you’ve encountered where the the FCC model fit plot is blank (and everything downstream of that may be nonsensical) is a known issue and will be fixed in the future. It seems to be a property of specific datasets - notice how the last column in the FCC plot is also blank. I’ll respond again when the fix is available.
– Harris
Not sure it’s specific datasets, but specific things done in earlier stages of processing. I see this effect in RBMC if I sample EER data at 2 (8K) but Fourier crop to 4K during motion correction.
Not sure if related but another avenue to investigate.
