Subparticle Extraction and Refinement from Icosahedral Particle in Cryosparc

Local Refinement
1

Hello everyone!

Is there a way to accomplish subparticle extraction and refinement from an icosahedral capsid CryoSPARC?

I am working on solving the structure of a 35 nm, T=3 icosahedral capsid and I am running into a resolution limit of exactly 4.37 angstroms. No matter what particle curation I do, I cannot get past this resolution. If I use all of the 8,000 auto-picked particles, I get 4.37 angstroms after refinement. If I do a bunch of rounds of 2D classification to remove all but the 1,300 best particles, I get 4.37 angstroms after refinement. The data was collected on a Glacios (200 kV) equipped with a Falcon 4 camera at a pixel size of 0.97 angstroms, so this resolution limit feels odd. I suspect it may be due to particle heterogeneity, as particle flexibility might be causing the loss of resolution during icosahedral averaging.

I’ve seen other research that got around this issue by doing icosahedral subparticle extraction and solving the pentavalent capsomere subparticle structure, but they used software external to CryoSPARC to achieve this (Cryo EM Analysis Reveals Inherent Flexibility of Authentic Murine Papillomavirus Capsids - PMC). I’ve been trying to come up with a workflow in CryoSPARC that would extract some kind of icosahedral subparticle (i.e., asymmetric unit, hexavalent/pentavalent subparticles), then do local refinement on those subparticles, but haven’t had any luck from my own reading and efforts. I included an image of my workflow with the volumes blurred below, as well as a couple images of the volume output at positive and negative thresholds.

image
image1484×1016 179 KB

Volume from local refinement at threshold of 0.15:

Screenshot 2026-05-29 114006
Screenshot 2026-05-29 1140061029×730 148 KB

Volume of local refinement at threshold of -0.09:

image
image680×538 43.4 KB

Are there any suggestions or obvious issues with this workflow?

Thank you for your time!

- Clark

2

Why are you inverting your mask?

This should be reasonably straightforward - the only step you might want to add is to use volume alignment tools to recenter the asymmetric unit and align the local symmetry axes, and enforce any local synmmetry (e.g. C5 if a pentamer) during local refinement.

Cheers

Oli

3

Easiest way I’ve found is as follows:

  1. Symmetry expand to C1 from I1
  2. Download map, use Chimera(X) volume tool, move to point of interest, erase outside sphere (50-100 usually works fairly well. Save volume erased volume.
  3. Import to CryoSPARC. Generate mask (do not invert).
  4. Use mask to set recentring point using Volume Alignment Tools on centre of mask density for volume and particles.
  5. Re-extract recentred particles.
  6. Homogeneous reconstruction.
  7. Local refinement (with recentring and sigma limits to mimic RELION local searches), then Local CTF refinement for larger particles, then another Local refinement if necessary.
  8. 3D classification into 12-30 classes, depending on point of focus, potential heterogeneity, etc.
  9. Continue local refinements, etc.

edit: This is 99% overkill for small particles like T=3 viruses.

1 Like
4

Also just to add - if it is a large particle, may be worth adjusting subparticle defocus using this script: