Hi, I used cryosparc v3.3.1 to perform 2D Class job. But the result shows some strange 2D averaging I didn’t see before (attatched picture below). There are some rings around the particles. Is it possible the pixel size or defocus is not accurate or other issue? Thanks!
Hi,
I am not the most experienced here, but, in my opinion, you need more particles there. The noise is still very high because the classes don’t have enough convergent signal. You can try putting all that in one or two classes just for visual, but I don’t know what you’ll be able to do with that. From my datasets, the nice-looking classes have ~1000 particles at least.
It looks like your box size is quite large compared to your particle size, so what you’re seeing a rotational averaging of background and neighboring particles around the actual particle. If you are worried about variation in size and position, you can play with the “Window” size to softly mask out neighboring density and even tighten up your mask, otherwise, I would recommend re-extracting with a smaller box size. From there, I would also re-run (and keep a log) of varying the number of classes and the initial uncertainty factor. A higher initial uncertainty factor will also produce these rings as it tends to distribute particles to classes with lower weighting and broader class distribution, but this can be balanced with more classes or a lower initial uncertainty and fewer classes depending on your heterogeneity in your sample conformation, population, and oligomerization.
All in all, I would say this isnt necessarily something to worry about unless you see artefacts in your 3D runs. Remember 2D classes are not necessarily reflective of 3D reconstruction ability or quality.
Hope this helps!
There are also some nice classes (100-500 particles/classes) in the same jobs. It’s just strange that some classes looks like this.
@jianhaoc The appearance of class averages may also be affected by Force Max over poses/shifts, which defaults to off when 20 or fewer classes are specified for classification.
1 Like