Hi,
I have a problem with preferential orientation in my collections, since my complex have an elongated shape (so I don’t have enough data from the top and the bottom of the complex) and also I have a preferential orientation even in the elongated position.
The data is high-resolution, high contrast, etc. But I’m wondering how to overcome that orientation distribution problem. Is it possible that collecting in tilt could be helpful? And if yes, would you recommend me tilt series or fixed angle? In that case, what angle(s) would be the best considering defocus estimation and ctf correction downstream?
Thanks 
1 Like
Hi, I’m also interested in this topic if anybody has experience with this. Also, what troubles one could have in downstream processing of a dataset with a fix tilt angle. Many thanks !
Tilting certainly works well in some cases, although there is an unavoidable tradeoff - as you increase the tilt, your effective ice thickness becomes thicker, and the magnitude of effective beam induced motion in the X-Y plane becomes larger, limiting resolution.
Sometimes it is the only way though! No need to do a tilt series. Can try say 30-40degree tilt to start with. See here for details:
Cheers
Oli
1 Like
In terms of downstream troubles, the main thing is to do Patch CTF from the get go, and then follow up with local ctf refinement after an initial consensus refinement of all good particles
Oli
I would second @olibclarke’s suggestions - my personal experience is limited but we have seen many examples of tilted collection that substantially improve orientation distribution-limited cases.
You should use Patch Motion (which has been tested heavily against tilted data at 40-60 degrees, where motion blur due to sample deformation is dramatically increased) and Patch CTF (which will be able to automatically detect the tilt/bend in the ice without knowing the tilt or direction ahead of time). You can actually combine tilted and untilted data collection in the same session as well, in case the untilted data can give better high-resolution details (in some directions) and the titled data can help get enough viewing directions for a feasible 3D reconstruction.
3 Likes
This is mentioned in the reference Oli posted above, but when you collect tilted data you almost certainly want to use a holey gold grid (aka gold-on-gold aka Ultraufoil).
2 Likes
Thanks @apunjani @olibclarke, this is extremely helpful to know how to work with a tilted dataset.
Would it be correct to merge selected classes from tilted and untilted data to make a new ab-initio, and proceed from there ? Otherwise I already have a refined volume from the untilted data, but it suffers from missing information.
Hi @marino-j,
Definitely it’s a good idea to start from ab-initio reconstruction again with the combination of the data, just to be more sure that the ab-initio structure you got before with untilted data is correct. If you are sure of that though, you can just take the new particles (tilted) and refine along with the untilted particles against the existing initial map, or even from the current refined map.
Note that in this case it’s worth turning on
Minimize over per-particle scale (in all refinements) as well as Use scales from current alignment in reconstruction (in Legacy refinement and NU-refinement) because your particles are likely to have different contrast between the tilted and untilted data. This will estimate per-particle scaling factors and help to make sure all the data gets used appropriately.
3 Likes
@apunjani many thanks for the suggestions ! Best wishes