Does Local refinement take account of higher order CTF parameters?

Local Refinement
1

Hi,

Does Local refinement (with NU-refine on) use tilt/trefoil/tetrafoil aberration corrections estimated using global CTF refinement? I know NU by default does not, but I am not sure what flavour of NU-refine is used in Local refinement under the hood. My guess is no but just want to confirm

Cheers
Oli

2

just bumping this in case anyone knows

3

Hi @olibclarke - local refinement does not yet use or correct for CTF aberrations (tilt/trefoil/etc) though it will respect local defocus values if those have been refined

4

Ok thanks Ali - good to know! I hope it does in a future version, as in particular taking account of beam tilt can make a big difference at high resolution for certain datasets.

Cheers
Oli

5

Hi @apunjani does local refinement incorporate support for higher order CTF parameters in v3?

At very high resolution it becomes essential, meaning we can’t use cryosparc for local refinement for high resolution datasets with v2.x. Is there a possibility of this being incorporated soon if it isn’t already?

6

Hi @olibclarke,

Unfortunately not yet – the local refinement implementation in v3.0.1 is the same as v2 aside from the bug fixes related to masking and other parameters, so it doesn’t respect the beam tilt or trefoil etc. It is currently being worked on, and we’re hoping to get this change out as soon as possible so that higher order CTF parameters are respected.

Best,
Michael

1 Like
7

Thanks very much Michael - good to know!

Oli

8

I does look like reconstruction uses higher-order CTF parameters, so you can at least follow up local refinement with a new reconstruction using the refined parameters.

2 Likes
9

That is good to know, thanks @DanielAsarnow!

10

Hi all,

@DanielAsarnow’s suggestion should definitely work in the case when the fulcrum is not specified (i.e. the local refinement was done with default fulcrum as the box center).

For local refinements with a specified fulcrum (not the box center), I have encountered a bug where the output alignments are incorrect – they don’t correspond to the alignments of the output volume due to the volume shifting done. This can be confirmed by running a homogeneous reconstruction job with the particles output from a local refinement with a non-zero fulcrum, so if you notice that the reconstruction looks incorrect in these cases, that would explain this error. Please note that this is only an issue with the alignments being written out incorrectly, and the severity of the error increases with the distance the fulcrum is from the box center. Unfortunately it may still cause problems if you want to do further processing on this stack (e.g. signal subtraction or 3d variability, etc.)

We’re hoping to release the updated version of local refinement that fixes this later this month (i.e. will avoid the issues caused with shifting the volumes, and includes support for new non-uniform refinement, marginalization, global symmetry, higher-order CTF, and per-iteration pose drift, etc.).

Best,
Michael