Plot in-plane rotation changes in Helical Refinement?

Helical Processing
1

Hi,

We have a helical case with a pseudosymmetry with respect to the directionality of the helix. The result is that after helical refinement, both directions are present in ~50/50 mix, distorting the map.

This can be fixed by a two-class classification without alignments after helical refinement, but during helical refinement it would perhaps be useful if in-plane rotation changes were plotted (rather than just tilt angle) to get a sense of whether it is still jumping around at the end of refinement.

Also, what does cryoSPARC do for searching the in plane rotation? I think in relion there is a bimodal -180/+180 gaussian prior, is that the case here too?

Might it be useful to have an option to “lock” in plane rotations for cases like this, where directionality is ambiguous, so that if one has corrected it by 3D classification it is possible to run another round of helical refinement without particles flipping?

Cheers
Oli

2

Hi @olibclarke,

Thanks for the request – this makes sense! Right now in helical refinement the plots show the absolute tilt angle, and shift along/perpendicular to the helical axis. But plotting the in-plane rotation change histogram would make sense, for ideal cases it would peak at 0 as iterations progress but for cases where the polarity isn’t very strong, it might have dual peaks near 0 and 180. I’ll definitely record this!

In helical refinement, CryoSPARC currently only modifies the pose search grid by excluding poses that are tilted out of plane greater than the provided parameter. All in-plane orientations are searched and there is no prior favouring any of them.

Have you found cases where after correcting by 3D classification (perhaps by rotating one of the classes by 180Âş to match the other, or using Align 3D Volumes), that a subsequent helical refinement using the reference from 3D classification, gets the polarities confused again?

Best,
Michael

3

Hi @mmclean - yes we have a case where the polarity is scrambled in helical refinement, even if it is correct to begin with… that is part of the motivation for the other request for enforcement of helical symmetry during local refinement :slight_smile:

1 Like
4

Hi @olibclarke,

Thanks, I’ve also recorded the suggestion to add an option to freeze the in-plane psi angles in helical refinement :slight_smile:. The only other parameter I could think of that might help alleviate this is changing the initial lowpass resolution value to something higher res, but this only helps if the polarity detail in the 3D density is recovered totally after the first iteration, which may be over-optimistic.

Best,
Michael

1 Like