I’m interested in finding a way to visualize 2D class averages directly on the raw particle images or their parent micrographs ideally to see how specific 2D features correspond to their original positions and orientations in the dataset.
The motivation is to better understand heterogeneity, orientation bias, and particle distribution within my data. By overlaying the 2D averages, I’d like to visualise where distinct views or conformations occur and how they relate to local signal, backgrounds in the micrograph. This could also serve as a useful diagnostic for assessing particle alignment and classification outcomes.
Is there an easy way to begin doing this using cryosparc-tools?
Thanks,
-J
As @carlos mentioned, it might be easiest to select a single 2D-class and connect those particles to an inspect picks or curate exposures job for easy visualization. Curate exposures offers a nice way to see all of the stats in plots for micrographs that contain particles. By loading in a single class of particles and setting the micrograph selection criteria to mics that have a minimum of 1 particle (or some other number of particles), you can look for any correlations between number of particles and other micrograph metrics such as ice thickness, defocus, etc.
For example of an ion channel I am working on that has some preferred orientation (missing side views, so much so I have a single 2D-class of side views). By connecting the particles from this select 2D-job to a curate exposures job, it appears there isn’t a strong correlation to micrograph level parameters from CTF estimation or motion correction. Surprisingly, there seems to be a cluster of micrographs based on their index number that seem to indicate there is a specific area of the grid where these views end up being more prevalent. I hope this helps, let me know if you have any other questions.
That’s super helpful, thank you. One of the really useful test cases I see for this approach is that it allows filtering of complexes and then performing single-molecule-style analyses — for example, measuring inter-particle distances or angles, and orientations. This doesn’t necessarily need to be done with cryo-EM data; even negative stain would work well for exploring this.
Is there an easy way to write out the micrographs with the 2D averages overlaid after the inspect picks? I know RELION implements something similar via particle_reposition, but I’m hoping to avoid jumping between packages if possible.
Thanks for the additional info, we have noted this request. If you are using inspect picks/curate exposures to display particles from a single class, there is a button to download the current micrograph. This would be rather laborious for a large number of micrographs though.
If you venture into using cryosparc-tools, you could display and and save every micrograph in a programmatic fashion.