since CryoSPARC currently lacks a jobtype for focused classification (i.e. masked heterogeneous refinement) I wonder if 3D Variability analysis with a local mask could be used instead. I tried 3DVA with a local mask to classify tRNA and antibiotics in ribosome datasets. The “cluster” mode in the 3DVA-display job seems to do a decent job in sorting particles and the corresponding volumes look reasonable.
Any thoughts on this strategy?
Thanks for your answers!
Hi @mannda hopefully this question is still relevant to your project - I would say that if you find that 3DVA is able to cluster or otherwise separate your particles based on conformational differences within a masked region, this is a great strategy. It is likely that 3DVA can find conformational heterogeneity with more sensitivity than a focussed 3D classification method (which would inherently be discrete).
However, you should make sure to perform a gold-standard homogeneous refinement (or non-uniform or local refinement) on any subsets of particles that you identify in this way, so you can be sure that any high-resolution signal you see appears in independent half-sets of the data.