Does the new ctfrefine function depend on the box size? I’ve tested it on several datasets, it appears that for the same dataset, a larger box size would provide a more reliable result. Is this right?
If this is the case, it might not be very efficient to re-extract a large box size, is it possible to perform this fitting with the corrected micrograph only utilizes the coordinate information from the refined particles.
Let me know if the problem is not described clearly.
@wxh180 Thanks for posting - yes, the ability to detect and refine CTF parameters per-particle (and across a group) does improve with larger box sizes, though there are probably marginal gains after a certain point (maybe 3x the particle diameter?)
The idea you mentioned, of doing CTF refinement against the micrographs rather than extracted huge particles is a great idea. We are actually planning something that’s more like
Extract large boxes -> on-the-fly cropping to smaller box size during refinement -> on-the-fly CTF refinement against the larger box.
This is still in the works.