Hi all, I had a problem and question about EER format movies on cryoSPARC patch-motion correction.
Recently, I collected data from a Krios microscope equipped with a Falcon-4i detector and obtained a gain reference file (.gain format file) from the Institute.
Then, I imported my movies into cryoSPARC (v4.2.1) with gain reference (no upsampling and final frame = 40), and I didn’t observe the black lines that are commonly reported with EER format files. However, I noticed that the gain reference file contain a white column of dotted lines, which may indicate defect points or zero signal points ? (Fig 1)
Regardless, the gain-normalized movies look good. I’m planning to proceed with Patch-Motion → blob pick → extract → 2D class in general.
However, I have encountered some peculiar 2D classes and their corresponding locations that I’ve never seen in my other cases using .mrc or .tiff formats.
These seem to align with the white lines present in the gain reference (Fig.2)
Fortunately, the Relion (v4) MotionCorr does not exhibit these peculiar 2D classes. Are there any significant differences in how cryoSPARC and Relion process EER format data?
I usually prefer cryoSPARC processing on my system, but in this situation on EER, I have to use Relion
Has anyone else encountered this problem?
I attempted to generate a gain reference from images (~ 250 movies) using ‘relion_estimate_gain’, but encountered the same problem.
I converted the .EER file to TIFF format using ‘relion_convert_to_tiff’, with both the original and generated gain reference, but the problem persisted.
Yes. I’ve explored as many combinations of flips and rotations as possible. If the wrong gain flips or rotations were chosen, the quality of the movies was significantly degraded (ex. additional appearance of black lines or the loss of particle features due to the highly intense signal from the white defect line)
Sure! each .eer movie is ~ 700 MB in size, so 100 EER files would total around 70 GB. What would be an efficient way to share these files with the Structura team?
Other group’s data collection from the same beamline Krios have reported similar issues, but it’s not certain that this issue occurs with all Falcon-4i detectors since I only have experience with one. It could be a problem specific to the beamline detector.
Also, I didn’t receive a ‘Defect file’ from the Institute. Is it possible the ‘Defect file’ could be generating directly from the Falcon-4i/EPU or .eer files? I’ve never used a defect file during movie import.
I am hopeful to identify what might be causing these issues.
Thank you.
Is the main issue that you’re experiencing odd concentric-ring 2D classes, coming from a particular region of the micrograph? Yes, I can clearly see these rings on the imported micrograph after motion correction, and they are also visible in the 2D classes. I can see protein particles “polluted” by these fringes. The rings are located along vertical lines, which can appear either on the right, the left, or on both sides of the micrograph. However, the features are not always located at the same position along these lines, and their number can vary between micrographs.
Does your gain reference indicate defects? No, not that I can see when visually inspecting the gain reference file.
Are you using a separate defect file in addition to a gain reference? No.
What format is your gain reference file? (.gain / .tiff / .mrc) .gain, as provided by our EM facility
As a temporary workaround, you could extract particles in relion, which has the option to remove pixels above or below a certain standard deviation from the mean.
You can also try using remove_outlier_pixels from CisTEM, which does more or less the same but for the micrographs.
Thanks for the tools, I’ll give them a try with relion.
That being said, it’s not a major issue in the sense that I can get rid of these artifacts after a few rounds of 2D/3D sorting. I’m just curious to understand the source of the problem and whether there’s something I can correct for my next data collection and/or processing habits.
@Syrup, I can’t be totally sure but it does sound like the issue is that there is some type of defect that isn’t accounted for in the gain file. CryoSPARC does remove outlier pixels greater than 5 standard deviations from the mean, but the threshold isn’t configurable. Oli’s suggestions may offer a workaround. As an alternative to those, if you can figure out where the defective pixel lines are, inserting zeros into the gain reference in those positions will cause CryoSPARC to treat the pixels along the defect line as defective. You’d have to re-import the movies though, and re-motion correct them, etc.
–Harris
