Noisy 2D classaverages


we try to classify a relatively small protein (100 kDa) in 2d and we get rather noisy results. It looks quite much like artifacts. Does anyone know were these could come from and what we could do to get better class averages?




Hi @david.haselbach, can you try turning off the option Force Max over poses/shifts ? this causes 2D calssificaiton to marginalize over poses and helps with small/low SNR datastets

1 Like

In addition to marginalization, I like to use 40 iterations of online-EM, batch size per class 200, and limit the alignment resolution to say 8A.

@DanielAsarnow and @apunjani indeed looks much better thanks for your suggestions.