Recentering helical segments


When running 2D classification on filaments, there is an option to align the filament axis vertically.

I wonder whether it might be worth adding an option to discard translations along the helical axis? This would allow for extraction with recentering for helical segments, whereas currently recentering will result in a lot of duplicate particles, particularly if the helix has a small rise.


Hi @olibclarke,

This makes sense, and this functionality is present in symmetry-enforced helical refinement (the “Limit shifts along the helical axis” parameter). After activating this and running a symmetry-enforced helical refinement, you should be able to extract with re-centering without generating a whole bunch of duplicates.

At the 2D classification stage, there are a few complications that make it awkward to do this – namely, we would need somewhat precise knowledge of the helical symmetry parameters (chiefly the rise), so that it can search over the asymmetric unit only. As well, there are some cases where estimating the filament axis during 2D classification fails (it is based on a heuristic, rather than any specific knowledge of the helical symmetry axes like in 3D), and so limiting search to be only over the ASU would break. Similarly, sometimes 2D classes of filaments closely correspond to 2D projections of the helix with nonzero out-of-plane tilt, which also would complicate the definition of the ASU in 2D…

Are there situations where you aren’t able to use the first method (via re-centering after 3D) and would still benefit from this feature after 2D classification?



Hi Michael - that’s reasonable, and I think you are right that recentering after 3D will suffice in most cases. Thanks!