How to refine pixel size?

Dear all,

I am currently in the later stages of refinement and my FSC curve looks bumby. Looking at my structure it might be that the pixel size is slightly off.

Is there an easy way to refine the pixel size in Cryosparc?



Hi @stdegiet,

Do you have an estimate of the correct pixel size from looking at the structure?
Unfortunately we have not yet implemented a method to automatically refine/optimize the magnification/pixel size.

Hey Ali,

Thanks for the response!

I do have an estimate-ish of the correct pixel size. What would be the best way to change it manually? Or is it better to re-run from scratch using a new pixel?



You can import the particles again and manually override the pixel size. However, it’s important that you refit the CTF with the correct pixel size either by re-running CTFFIND on the micrographs or performing defocus refinement. The pixel size has a strong effect computation of the CTF.

Also, in my experience, pixel size errors look different than that, with a more pronounced dip closer to 4 A followed by a relatively sharp peak (i.e. crosses 0.143 twice). What you have looks more similar to the membrane protein dip, which I believe is caused by persistent misalignment. Definitely worth giving it a test anyway.

Dear Daniel. Thank you for the response.

Indeed, I am working with a membrane protein so the dip might be related to that.

I will rerun the Ctf and check if it improves the curve at the different pixel size.


Sorry to revive this old thread but are there any updates to implementing this?

Anisotropic magnification can be refined in global CTF refinement now. The best way to determine the true pixel size though is to use a very high resolution refinement and calculate it from refined Cs values.

1 Like