Advice for 3D classification of noisy, flexible region

Hi all,

I’m trying to perform a focused 3D classification on a small, flexible “head” domain of my protein, which can rotate and nod relative to the “body” of the protein. However, while I’ve been able to resolve the body to a resolution of 2.6 A, the density for the head domain is extremely noisy - it looks more like a really dense region of dust instead of something solid. I have a few questions:

What’s the best way to generate a solvent mask that includes the head domain? The mask from the upstream refinement job completely excludes the focus mask on the head domain. If I’m understanding the solvent mask section in the tutorial correctly, this means that the density inside the focus mask will be completely ignored. What would be the best way of resolving this? Lowering the “auto solvent mask threshold” to the point where it includes the head domain ends up including a lot of noise throughout the box. Right now, I’m just combining the solvent mask from the upstream refinement job with the focus mask using volume maximum in ChimeraX. Is there a way to do this type of mask combination in cryoSPARC?

How soft does a mask need to be for density with really poor resolution? The mask creation tutorial recommends a minimum soft padding of 5 * resolution / pixel size, and I assume the resolution refers to the resolution inside the mask. However, this seems like it would result in an absurdly large soft edge if the resolution is 10A or worse.

Cheers,
cbeck