Box size in particle extraction vs. refinement

I have a question about how box sizes translate across jobs in cryosparc. I originally extracted my particles at 400px, bin 4 (cropped to 100px) for ab-initio and hetergeneous refinement. Once I had my good particle set, I re-extracted at 400px and did NU refinement. When examined this final particle set by 2D classification, I noticed that the edge of the particle seems slightly cut off by the mask (and resolution at the edge of my particle is also lower), so I want to re-extract the good particles at a larger box size (500px?) and repeat the NU refinement.

Now, my question is: can I use my previous volume from my NU refine job at 400px as the initial model for my new NU refine job? Or do I need to re-do ab-initio and hetero refinement using the 500px particles to get an appropriately scaled volume for NU refine? Or is there a way to adapt my current NU refine volume for use with particles extracted at 500px?

It would be great if someone can help me understand how pixel sizes are used across jobs.

You could change the default windowing from 0.85 to 0.9 or so, and see if that makes a difference, but if your box is that tight you might be leaving resolution on the table (depending on your resolution of reconstruction).

I would re-extract, use the current NU refine result and run again, it’ll be OK as filtered to 30 Å, the map loses all features but the most general shape anyway.

If it’s caused by the mask, I would suggest autotightening is being too enthusiastic, but if it’s a soft falloff around the extremes of the box, then it definitely sounds like the windowing function is the culprit.

Also, generally, it’s recommended to have a box size which is ~2-2.5 tiles the size of your particle. If you have particles close to the edge of the micrograph, this will mean you will lose a lot of particles upon reextraction into the larger box as CryoSPARC will not extract particles where the box goes over the edge of the micrograph.

@strawb,

At the beginning of any kind of refinement, CryoSparc will rescale the volume under the hood (thanks CS crew for this!), according to the pixel size of the particle set, so the initial voxel size doesn’t matter. This is usually stated in the logs, it is of good practice to check them as there might be important information. I usually spot my mistakes like this, during the first few minutes of job running.

Luck,

1 Like