Converting cryoSPARC refinement back to Relion to do 3D classification without image alignment

Hi all,

I want to import my cryoSPARC coordinates to Relion and do 3D classification without image alignment (because as far as I can tell, I can’t do that in cryoSPARC). I also want to do masked multi-reference 3D classification (which isn’t yet supported in cryoSPARC).

When I use @DanielAsarnow 's helpful script ( to do the conversion, my Relion refinements always come out odd. Typically they take 5x longer to calculate and the resulting reconstructions just look odd and stay in 10 Å resolution range (I get 2.8 Å in cryosparc).

I’ve seen that others experience this issue: Convert csv file to star file and keep orientation parameters @olibclarke @Rafael-Ayala

Has anyone recently made progress on this issue, or figured out why the Relion refinements turn out odd?


@sjweaver Thanks for trying my program! I guess I missed the other thread back then. I just updated the tutorial on the pyem github wiki. The most up-to-date program and documentation can always be found in that repository.

At the end of the new tutorial, there are some detailed notes about why the results may not be as you expect. Let me know if you think something is missing - I suspect your issue is related to one (or more) of the suggestions there.

Thanks @DanielAsarnow ! I’m probably making a mistake somewhere. I’ll have a look at the new information and try it again. I’m glad to hear (in the other thread – linking for people in the future to find things easily Convert csv file to star file and keep orientation parameters ) that it is working well for everyone now.

Thanks for including more information in your wiki page. It’s very helpful to those of us who are new to cryoSPARC.