Domains getting eliminated or washed out with NU Refine

I have a C3 symmetric molecule. The map gets worse as you move out radially from the core.
I did ab initio-> het refine->homo refine to get to ~3 A res. A subsequent NU-Refine job pushes res to 2.8 but also removes the outermost density in the process (Fab antibody). The domain adjacent to the Fab gets totally washed out/density gets significantly worse.
Is it better in this instance to skip NU refine and go straight from homo refine to local refinement? I am relatively new to this- appreciate any advice.

I would perform various methods of conformational sorting before local refinement. 3D variability in cluster mode, 3D classification with many classes, more rounds of het refine. It’s hard for local refine to align something that is so mobile it is unresolved with high-resolution refinement of an aligned subparticle. Can also try partial signal subtraction with NU-refine of the Fab if that’s what’s of interest.

Fasttrack suggestion: take all particles from NU-refine, run 3D class filter resolution 8 number of classes 10 or 20 or 30 (or more) such that >5k particles per class, yes keep data at each F-EM iteration. I would expect most of the volumes would have same core structure and FAb in different locations. Take subclasses and refine.

I ignored all symmetry aspects, but with more info it’s possible to use that to your benefit as well, at slightly increased complexity.

1 Like

thank you! I will do more sorting and report back.