Different pixel size in particles .star file and initial .mrc model in Relion job after conversion

Good afternoon,

I have a question related to the import files from cryoSPARC to Relion3.
I have followed instruction from the page Export from cryoSPARC v2 and converted my parlicles.cs file to particles.star.

command: csparc2star.py my_file.cs my_file.star

I used an ab initio model (.mrc map) generated by cryoSPARC, as a reference and converted .star file as an input for the class3D job in Relion but it gives me an error message:

ERROR The reference pixel size is 2.108 A/px, but the pixel size of the first optics group of the data is 1.054 A/px!
The reference box size is 128 px, but the box size of the first optics group of the data is 256 px!

If I run Relion 3D classification job with a .mrc map generated from cryoSPARC refinement job instead everything is working fine without an error.

Could you please explain why this maps has different parameters (pixel size and box size)?
Or if they are the same, why in my Relion jobs it recognized differently?

Thank you.
Best regards.


this is because cryosparc has internally downsampled your particles to speed up ab initio. By default ab initio is done in a 128px box. Is there a particular reason why you want the particles from ab initio rather than refinement?

Thank you very much for the answer.
I am going to repeat 3D classification and 3D refinement in Relion to compare my results, and decided to use my ab initio model instead of refined model since I though that this might reduce a bias for resolving some flexible elements.

Apart from that I do not have any other reason and jobs are running with refined model as a reference as I mentioned before and particles parameters correspond to the model.

Thank you very much again since I was a bit confused about myself unintentionally changing settings (binning or resolution) at some stage of my workflow.



You just need to scale the map back to the right size:
relion_image_handler --i abinitio.mrc --o abinitio_256px --new_box 256 --rescale_angpix 1.054

You can use the .star from the ab initio job if it’s convenient, just remember you probably don’t want to use the actual angle/shift alignments for anything. (No local searches).

Thank very much Daniel for your suggestion, this command would be very handy.