Increased Epochs for Micrograph Denoiser

Cryo-EM Data Processing
1

Hi,

Does anyone have advice for how much to increase the number of training epochs during micrograph denoising? I see it is recommended to increase the number for low SNR particles.

I have a small protein complex, ~130 kDa total mass and ~150 A in its longest dimension.

Denoising definitely appears to help for my low SNR dataset, I’m just curious how good it can be. Is it just an iterative process? Or is there a sort of maximum number of epochs?

Thanks!

David

2

On a couple of samples (<90kDa) I saw a big improvement for picking with 400 iterations. For some larger complexes with very different contrasts at different angles (large, flat membrane protein), I found 300 worked well, but 400 denoised out the low contrast views.

Now I routinely just set 300 epochs, as it’s not that much slower than the 200 default but works nearly as well as 400 for smaller particles.

I’ve not tried higher than 400.

2 Likes