Weird behavior when doing NU refinement with particles previously aligned to a different reference


I’ve had repeated issues with NU refinement form particles previously aligned to a different reference, either an ab-initio reference with a different orientation or a flipped reference. It looks like the particles don’t overcome the prior information. As a result, many particles remain misaligned. It is possible that they don’t have a lot of features to align. (It’s a small MB protein complex. )
Doing a 2D before 3D refinement resolves the issue.

I can send a picture of the resulting volume offline if needed.


Hi Amedee, do you mean a local refinement with NU activated?

All the other modes homogeneous, non-uniform, and heterogeneous refinement start from de novo global alignment searches.

Hi @adesgeorges,

@DanielAsarnow is correct that all cryoSPARC refinement jobs other than local refinement and 3D variability start from scratch for particle alignments, so I think this must either be a coincidence (due to stochastic differences between multiple runs of non-uniform refinement) or very strange bug. Could you DM me or email with more information/images?