I am very new to cryosparc but have some promising results from non-uniform refine. The protein core is more rigid than the surface due to many flexible glycans surrounding it. I am wondering whether the B-factor sharpening performed by NU-refine is at a fixed value or does it vary throughout the structure? Is there a way in cryosparc to apply more sharpening to the core and less to the glycans? At the moment the glycan density is reduced to a stump in the volume_map_sharp mrc from NU-refine while the output volume from heterogeneous refine shows low resolution features at glycosylation sites. Also, which parameters are usually best to optimize during NU-refine? I have used the default.
Great to hear you have good results!
All B-factor sharpening currently in cryoSPARC is global i.e. the same across the map. Typically users will create multiple sharpened maps (using the “Sharpening” job type multiple times, after NU-refinement) to create maps with different sharpening where different regions are better visible.
Parameters that can be used to optimize NU-refinement (presuming you are using the “Nonuniform refinement new” job type in v3.0+) are mainly:
- Number of extra final passes : can set to 2 or more to see if resolution may (slightly) improve after initial convergence
- Non-uniform AWF : see the paper in Nature Methods for more details about this, it can sometimes make a difference but also not huge
- Minimize over per-particle scale: if there is variance in ice thickness across the dataset, this can help
- Optimize per-particle defocus : have to empirically try to see if you have enough SNR for this but it can help a lot
- Optimize per-group CTF params : have to empirically try to see if you have enough SNR for this but it can help a lot
Thanks for the reply @apunjani. I will look into the B-factor sharpening but first want to check something with you regarding my NU-refinement output.
The glycans have a strange shape when I view the unsharpened half-maps and I am wondering if something went wrong during refinement? It is a core with some density surrounding it or branched. I used the default parameters for dynamic masking. Do you have any idea why it looks like this? Or is this normal? Should I change the masking in any way?
I switched global ctf refinement, defocus refinement, trefoil, tilt and tetra off and used particles and volume from one hetero refine class as input.
Here are some output images: