Class probability filter output for heterogeneous refinement

Hi all,

I am attempting to use the “Class Probability Filter” job to sort particles by their posterior probability of assignment to various classes of a Heterogeneous Refinement job. When I run the job, everything appears to work fine, but one of the lines in the output says:

[CPU: 823.7 MB] Selecting particles with total posterior probability >= 0.7 from Ab-Initio Reconstruction (calculated from sum of fields [‘alignments_class_0/class_posterior’, ‘alignments_class_1/class_posterior’])

Because it says “Ab-Initio Reconstruction” on that line, I am not sure if (1) the job is erroneously using passthrough Ab-Initio posteriors instead of the Heterogeneous Refinement posteriors (to be clear, the “particles” object that I used was the “particles_all_classes” object output by the Heterogeneous Refinement job), (2) the job is sorting on the requested posteriors but the output text is erroneous, or (3) I’m misunderstanding something here.

Please let me know if you have any insight here. If it’s Explanation #2, I’m good to go, but if it’s Explanation #1 I’d want to know about it!


Hi Josh - just looked in my log files and I see this too, even in the cases where there was no ab initio job previously run, so my guess is it falls into (2), but would be good to have confirmation of that. My suspicion is that this is a remnant of code ported from cryosparc v0, where the class probability filter was built into the mult-class ab initio job.


Thanks Oli! Your ab-initio-less pipelines are a pretty solid indication that it’s #2. Hopefully a cryoSPARC team member can confirm.

Hi all,

Explanation #2 is right, the “Ab-initio Reconstruction” is hard coded in there. The alignment field that is read from the particles should be the same as the input slot (labelled “required”) in the job builder. We’ll update the log to be clearer.