Is it possible to use particle sets to match particles before and after a RBMC job?

Import and Export
1

I have a particle stack that I exported into RELION for 3D classification without alignment. While that was running I ran a RBMC job in cryosparc with the same exported particle stack and got an increase in resolution. I would now like to reimport the different classes from the RELION 3D classification to re-refine in cryoSPARC. In order to avoid re-doing the RBMC I’m wondering if there is a way to match the particles in each of the classes reimported from RELION to the particles that came out of the RBMC job?

Normally when reimporting I would use a particle sets job where input A is the last cryosparc refinement job of the particles that were exported to RELION, input B is the particles from an import particle stacks job run with a Star file from RELION corresponding to one of my 3D classification classes as input, action is set to intersect, Field to intersect is set to path, and ignore leading UID is ticked. Intersection (keeping A data) then gives me the cryoSPARC particles corresponding to my RELION 3D classification class. This means I preserve the CTF information from cryoSPARC to prevent issues with the different format these are stored in between CS and RELION.

I am wondering if something similar is possible with the RBMC job, however the output particle paths change significantly in the output from the RBMC job. Is it possible to reassociate on another statistic, e.g. particle location? Or can the RBMC output particle file names be used to reassociate in some way with a script of some sort? The RBMC output files are of this form: 7364139056143776840_particles.mrc_0 while the original files are of this form: FoilHole_10318584_Data_10316533_4_20260310_170639_Fractions_patch_aligned_doseweighted_particles.mrc_10. Does the RBMC file number prefix relate to the original files in some way?

Thanks for your help!

2

can’t you still do the particle sets trick, then use the low-level input swap of RBMC CTF with each class in the refinement job?

3

I think I’ve found the answer to this. First use an intersection to get the particles from the original NUref (before RBMC) that correspond to the imported RELION class as described above. Then create another particle sets job, give it as input A the intersected particles originating from the NUref from the first particle sets job, for input B give the particles from RBMC. For parameters use Action: intersect, Field to intersect: uid, untick ignore uid. This then identifies the particles in the RBMC according to their uid, which is preserved in the name given to the particles by RBMC, even if the original filename is lost.