Empty images/2D projections - cleaning

Dear colleagues,

What is the way to remove the empty or badly picked particles?

Is there a way to inspect and clean the particle dataset manually, as it exists for the 2D class selection?

Kind regards,

Sounds like a fun afternoon (or decade)…if you have so many mispicks that throwing them away by 2D class doesn’t work, it might be more fruitful to tune the particle picking. In my experience, most users don’t spend enough time optimizing. You can set the number of micrographs to process to 10, in which case template and blob picking will both complete in seconds, and then iterate through different parameters until the results are good. For a challenging case it might take 30 tries, but they can be done in 15 minutes.

Also, don’t forget to use “inspect picks” and set thresholds that are inclusive on both high and low defocus images.


Hello @DanielAsarnow , thanks for your advice! Usually, I have a similar method working in CS. I was just wondering if there are currently more tools available and /or other development in particle picking within CS. Of course, it takes sometimes a lot of time, but perhaps you may agree that for some samples, manual picking/inspecting may be one of the successful strategies to proceed. Kind regards,

Actually, have you ever tried the eman2 neural net picker? It’s good for hard cases, and only requires ~100 examples. It was used in most the RNA-only structures so far for example. The GUI also affords the kind of per-pick inspection you are talking about.

(It’s implemented very differently from other neural network based pickers).

1 Like

hello Daniel,

No, I did not. I would like to - good idea. Thank you!