Hi, apologies again for the amateurish question. I am working with a particle that has a C15 symmetry, and I refined it to a 2.4 resolution using NU-refinement, followed by local refinement using a whole particle mask, so I am not yet doing any classification of any kind. Now I want to do 3DVA, and I am not sure whether or not to do symmetry expansion. I already enforced C15 symmetry during my refinements. I did the 3DVA with and without the symmetry expanded particles and I can see that the results in the “simple” mode output are completely different…so what is the correct way to do this?
If you enforced symmetry during refinement it is appropriate and acceptable to run symmetry expansion prior to 3DVA or 3D classification. This way, each protomer of each particle is aligned and it can effectively search over all the symmetry-related poses at once.
Interpreting the results can get tricky on whether you want to keep or remove duplicates when reconstructing or refining subsets after symmetry expansion.
You can also run refinement with symmetry relaxation, which will give a true C1 map, hopefully with effective sorting of asymmetric protomer differences, but you be the judge of that. Then run 3DVA on this.
Thanks a lot, so I should not use, say, the resultant clusters as is for further analysis? for example if I am interested in one cluster, it is not appropriate to use the resultant volume to create a mask and then refine a particular domain or region in this map?
I think what you describe is probably fine. I would generally not make a mask on the volume from 3DVA, but rather homogeneous reconstruct the particle cluster as an intermediate step in further analysis and make the mask on the reconstructed volume. The reconstructed volume can have different features from the direct volume output of 3DVA or 3D classification.
These days cryosparc will resample a mask with different voxels on grid of the particle reconstruction so they match (but it does not align them so be careful). Making the mask on the reconstructed volume used to be required.
So now if I have, say 5 clusters, I cannot directly use the output map of these clusters for anything other than visualization? but then how can I use the particles of each cluster as an input for homogenous refinement when I know they were symmetry expanded? for example the 15,000 particles of cluster 1 are actually 3000 because I have a C5 symmetry…so how can I get the original particles only?
I can use 3D classification right from the start for the discreet classes…but then if I do 3DVA on each of the discreet classes I will run into the same problem…
after symmetry expansion, you can use them for local refinement (let’s say you identified a “squished”, asymmetric structure, and want to refine that, this would be one way to go), just don’t put these clusters into NU-refine or homogeneous refinement, as global alignment can super-impose symmetry expanded copies and thereby lead to very funky FSCs.