Particle Sets Tool, intersect: which exact UID is used?

Hello,

A question came up during analysis of a dataset, but I don’t have access to the CryoSPARC instance hosting the project to run tests to answer the question. So asking here instead.

For context, here is what we have been doing:

  1. Homogeneous refinement of a symmetric particle, with symmetry; let’s call this set P for “parent particles”.
  2. Symmetry expansion to work at the level of the asymmetric unit.
  3. Masked 3D classification of all asymmetric units. We found classes with missing density likely due to partial denaturation at the air-water interface, let’s call this set D for “damaged” sub-particles. Let’s call the other classes I for “intact” sub-particles.
  4. Remove all parent whole-particles from set P that contain any number of damaged asymmetric units from set D.

For step 4, P inter D or P inter I based on UID should give us the intact subset of P (either as P-D or as P inter I) if the UID being considered for sets of sub-particles D and I is sym_expand/src_uid. Is this indeed the field considered by this job?

I think yes, because if the job were looking at the “top-level” UID for all particles and sub-particles, then these intersections would always yield an empty set (since U in UID means “unique”). And we don’t get empty sets. But the job’s interface only displays “UID” without specifying which exact UID is used for sub-particles obtained by symmetry expansion. The documentation doesn’t say either. I ran a quick test on one of my datasets that has the same sequence of symmetry expansion and 3D classification, and inspected the job log, but it also only says “UID” without being more specific. If someone from Structura Bio could confirm which exact UID is used, and update the documentation and job log, it would be very helpful. :folded_hands:

In addition to this, we get numbers of particles in the results of set operations that don’t perfectly add up (the discrepancy is small, but we expect set operations to add up perfectly), so we’re confused.

Thank you in advance!

Hi @Guillaume ,

not sure I got the full picture, but shouldn’t the analysis be performed on non-expanded particles? using the symm-expansion you can see the difference that are averaged out by real symmetry, but I thought it was not recommended to further down-process them (unless I’m mistaking).

What about running variability analysis in cluster mode on the parent particles to subdivide them?

It’s always tricky. Good luck.

Vincent

There is no analysis done downstream other than trying to exclude the damaged particles by set operations. For example, we already re-ran homogeneous refinement on whole particles after excluding particles based on this kind of intersection, and it worked fine. We simply would like to understand if we excluded the population we think we did, because the numbers from set operations don’t add up perfectly, and because of this ambiguity on which UID the intersection is based on.

1 Like

Linking some other relevant posts: