Empty 2D class averages with non-blank output images

[FabianKohler]

Hi, in a 2D classification of a negative stain dataset, I noticed classes which contain zero particles but show a non-blank image (see screenshot). Shouldn’t a a class without particles give a blank image?

The number of particles is correct (selecting a class containing 5 particles gives 5 particles, also when running a re-extraction job).
This made me wonder if maybe after the last iteration the class images are not updated and the images are actual the averages of the previous iteration… Did someone experience something similar and could comment on this please? Thank you for your help.

PS: For the cases where the class image looks blank but actually contains particles, it seems to me that an auto-brightness adjustment applied on the entire set of class images likely makes them appear blank in the overview. The mrc files of these classes are ok.

[FabianKohler]

Cryosparc Version was: v3.3.2+220824

[CryoEM1]

cool that you’re studying a tire rim! This is a normal and not particularly interesting phenomenon. I think it is due to the soft classification where particles in some cases belong in multiple classes with a percentage applied. so 40% in class 10 and 60% in class 11 for instance. at the end they’re obviously discretely “hard classified” but the images are likely remnants of early iterations.