Considerations for working with symmetry expanded particles in M

[thomas.la...@gmail.com]

Hi folks!

Curious if anyone has had the chance to use M on symmetry-expanded 2D data and what should be considered in doing so?

I have a pseudo-octahedral particle that ends up a mess when enforcing O, but refining sub-particles for the 6 faces with C4 seems to work well (enough).

I pre-processed in Warp and am working in RELION-3.1 using the sub-particles obtained by symmetry expansion, removing redundant insertions of the 4-folds, and re-centering on a 4-fold with the signal-subtraction job-type (may have boxed myself into a corner with this approach :P).

I've been attempting to perform Polishing, but this results in the reconstruction getting worse. Also tried substituting a run_data.star using the original particles reduced to the original particle count (i.e., each particle is represented once for Polishing), but got similarly poor results.

For 2D-data, would one expect M's reference-based motion refinement to get thrown-off with symmetry expanded particles, which appears to be the case for RELION's Bayesian Polishing?

Best,

Tom

[teg...@gmail.com]

Hi Tom,

I haven't tried it, but it should work in principle (unless you already have evidence to the contrary :-P). It will result in excessive memory consumption since there is no elegant mechanism implemented for reusing the same image for several particle instances. You should also make sure the same particle image doesn't contribute to both half-maps, otherwise there will be some crosstalk.

Cheers,

Dimitry

[thomas.la...@gmail.com]

No evidence one way or the other at the moment.

Will give it go and report back to the group.

Thanks for the insight!

Cheers,

-Tom

[thomas.la...@gmail.com]

OK, figured out where I went horribly wrong.

Sub-cropping particle images for re-centering was causing the havoc since I wasn't updating the particle positions in the original micrographs.

Addressed that issue and both Bayesian polishing and M yield improvements.

Cheers,

-Tom

[thomas.la...@gmail.com]

Hi again Dimitry and Warp/M community,

Re-opening this for a related question (hopefully it can serve as a reference in general).

I've found performing the symmetry expansion within M (after a symmetric refinement) is working a tad better for some reason.

After refining Image Warp, particle poses, and CTF (defocus & 3rd order) on the larger symmetric particle, I expanded to C1 through M and substituted a monomer mask in the species directory.

Then, I refined particle poses for the ~60 kDa sub-particle.

If I want to refine Image warp again (may be I shouldn't given the sub-particle size):

• Does M take care of the sub-particles sharing the same coordinates if I leave them unshifted?
• If I shift the particle center through the M GUI, do I need to apply these shifts to the references and mask in the species directories?
   

Refinement is taking 3-days, so would prefer not to "guess-and-check" but will do if necessary.

Sending other projects through M in the meantime for more great results (Mmmmmmmmmmm)!

Thank for the help,

-Tom