Option for integrated orientation diagnostics in heterogenous reconstruction?


The new orientation diagnostics job is extremely helpful, clear, and relatively fast to run. So far, it has the clearest and easiest to interpret diagnostics for anisotropy that I have seen.

Often during classification of a sample with preferred orientation, we are looking to identify a class not with improved global resolution, but instead with more isotropic resolution distribution.

To this end, it would be fantastic to have an option to integrate orientation diagnostics into heterogeneous reconstruction, so that orientation diagnostics is run on each class as it is being reconstructed. This would help greatly in identifying classes with more isotropic density.


EDIT: It would actually be great to have orientation diagnostics also automatically (default-on) incorporated into the final round of homogenous and non uniform refinement. It is something that not enough people pay attention to, and it is now fast enough that I think there’s not too much of a downside.

I would also display a warning when cFAR<0.5 (indicating possible anisotropy) and a more emphatic warning when cFAR<0.1 (indicating definite severe anisotropy, and advising caution in map interpretation).

In my experience cFAR is the most informative metric, as it relates directly to signal content, and is not as affected by how noisy the directional FSCs are (because it relates to area under the curve, not the point where the curve drops below some critical value). Turning this on by default would I think be a service to the community :sweat_smile:


Hey @olibclarke,

Thanks for the kind words, glad it’s been useful! We’ve noted the requests.


1 Like

@olibclarke – CryoSPARC v4.5 (released on May 7, 2024) now includes cFSC plots at every iteration of all refinements. Furthermore, the updated Orientation Diagnostics job can now accept a multi volume group from Hetero Reconstruct (and any other job that outputs it), and it will produce diagnostics for each class.

Thanks for your feedback!


Awesome!! Thanks Valentin!!