Refinement to enrich for complex particles instead of 2D classification

[ccgauvin94]

In general, if you can sort in 3D, I think 3D performs better. But if I’m understanding what got you to try it in 3D, it’s that you have so many particle types, and views, that 200 classes doesn’t give you any junk classes?

If so, you may be interested in the discussion in this thread, about the ‘anneal sigma’ setting in 2D classification:

Limiting the number of particles per 2D class

To follow up on this, here is an example (using the ryanodine receptor as a test case). Defaults: Start sigma annealing at it=2, for 15 iterations, 40 O-EM iterations, 5 full iterations: [Screenshot 2023-03-23 at 11.59.29 AM] All “bad” classes are essntially empty. No sigma annealing - Set sigma annealing start iteration to 200, effectively switching it off; other parameters unchanged: [Screenshot 2023-03-23 at 11.21.15 AM] There are now several thousand particles in bad classes. These p…

Regarding your hetero refinement question:

@ctueting

Unfortunately, Hetero Refinement does not seem to have an output “Unused particles”, like ab initio has.

I’m not sure I understand. Hetero-refinement will refine all the particles you tell it to. All of them must go to at least one volume, right? So “unused particles” would just be the particles in the classes you aren’t interested in, right? A lot of people will stop their ab-initio after the first iteration (when everything is an amorphous blob) to generate the “junk” map for their het refinement. Then, if you force hard classification, it’s pretty easy to use het refine like a filter.