I am having issues with some of my refinements due to flipping of either the map under refinement (or alignment of particles might be more precise) or the mask. To explain a bit more: In the first refinement job the mask and the volume to be refined are well aligned, but after global CTF refinement and creating a new non-uniform refinement, the mask has stayed on the same orientation, but the volume under refinement is upside down. I assume this can happen because the volume looks somehow similar in both directions and it does some kind or re-alignment between the mask and the volume? I have tried to find an option to avoid this, but did not find it. Anyone who has run into the same issue and knows how to fix it?
Hi @are! This may be happening when the volume is lowpass filtered during the first iteration of the refinement.
As I’m sure you’re aware, there’s not typically any way to tell the handedness of a particle image, so the volume has a 50/50 chance of having the wrong handedness. Once the resolution is high enough, we can tell by eye whether the handedness is right, and correct if not. Typically, subsequent volumes will retain the correct hand so the problem is a once-per-dataset issue.
By default, the first iteration of the refinement starts with a lowpass filter of 30 Å. Typically this leaves enough information about the volume hand such that particles align correctly. However, if the volume is symmetrical (or nearly so), a 30 Å lowpass filter may make the volume look so much like its mirror image that the particles align with the wrong hand again. This would result in a vertically-flipped volume as you describe.
To see if this is what’s happening, you could try decreasing the Initial lowpass filter (A) parameter in your refinement. 20 Å is typically still a very safe choice – you can go as low as a few Å greater than your GSFSC resolution and still be fine.
If anything here is unclear, or if the flipping issue persists, please do let me know!
Thanks for the reply. Yes, so it was not related to the handiness, but for the case of a rather symmetrical protein. Indeed changing the lowpass filter to 15 Å did the job in this case.