Focused classification of asymmetric features in icosahedral virus

Cryo-EM Data Processing
1

Hi all,

I’m new to cryoSparc and am wondering what the best workflow would be to carry out a focused classification to resolve asymmetric features in a virus? Specifically, let’s say that I want to improve the density of an external domain around a symmetry axis. Is there a way to do this workflow completely in cryoSparc?

Thanks!

Best,
Tobi

2

Hi Tobi,

It might be possible now that symmetry expansion exists, in conjunction with signal subtraction. E.g.refine with sym, sym expand, subtract everything but the desired region, then run multi class ab initio or heterogeneous refinement on the signal subtracted particles.

Personally I think this workflow is still better done partly in Relion for now though, as there is no way yet to perform a focused classification in cryosparc.

Cheers
Oli

3

Thanks Oli!

My workflow for now looks like this:

  1. Symmetry expansion of particles used for Homogeneous Refinement
  2. Prepare a mask in Chimera that encompasses everything except the density of one viral vertex around the 2-fold symmetry axis
  3. Run a particle subtraction job using this mask
    [The subtracted particles I get do look slightly different, i.e. the density for the viral shell is lower, however, there’s still lots of visible density, I don’t know why]
  4. Run a Local Refinement job using the expanded particle set + an inverse mask encompassing only the 2-fold vertex.

Now, I do get some density for the external domains around the 2-fold sym axis that looks better than what I got from a Homogeneious or Heterogeneous Refinement. What is strange though is that I do get a lot of other density (noise?) that is reconstructed. I thought that the mask would mean I would only get density inside the mask? Here is what I am talking about (we are looking down the 2-fold axis):

image
image1024×1024 321 KB

As you can see, there’s lots of density in this map outside the mask (which is basically the more well defined part around the 2-fold axis).

Any ideas on how to optimize this? Thanks.

Best,
Tobi

4

Hi @twg, @olibclarke thanks for discussing this workflow here - we have some work to do in cryoSPARC to properly support virus symmetries and local classification of this type. The more information/tips you have, the better!

Local refinement in cryoSPARC does use the mask during alignment of particles, but the mask is not applied again after reconstruction, so the maps you get out include density in the masked out region as well (but that density was never used to align the particles). The noise/mask-edge artefacts that can happen are a known issue and part of what needs work to improve on our side.

One thing that might be interesting to try out to resolve symmetry breaking features in subunits is 3D Variability. You can

  1. symmetry expand particles after refinement
  2. create a mask that includes only one subunit
  3. use the particles and mask in 3D Var to compute variability modes within that single subunit, using all subunits from all particles as data points

With that you might be able to see symmetry-breaking or pseudo-symmetry features more easily than with focussed classification or local refinement.

4 Likes