Ideal resolutions before performing CTF refinement and RBMC?

[cbeck]

Hi everyone!

I’m looking for some guidance on what resolution my map should ideally reach before trying CTF refinement and reference-based motion correction (RBMC).

From what I understand in the CTF refinement tutorial and a post from a few weeks ago,

CTF Refinement error in Non-uniform refinement

Hi @Izana! Welcome to the forum! Let’s see if we can clear a few things up about CTF refinement. The standalone Local CTF job Starting with the most straightforward problem, it looks like your Local CTF job has an empty half-set. This can occur when the particles come straight from an Ab initio reconstruction job. If this is true, first refining the particles (using, e.g., Homogeneous Refinement) or re-splitting them (by, e.g., using Homogeneous Reconstruction Only with Force re-do GS split tur…

local CTF refinement seems to need a resolution better than 4 Å, but I’m a bit unsure about the recommended resolution for global CTF refinement. The tutorial mentions the need for medium-to-high resolution signal, but I don’t have a good sense of what resolutions qualify. Should I aim for higher than 3 Å? And does RBMC also benefit from a resolution greater than 3-4 Å?

I’d also love to get a sense of where CTF refinement, RBMC, and optimizing per-particle scales best fit into a typical workflow. Do you usually perform these steps right after getting a high-resolution consensus, or are these typically the very last steps of data processing? Also, would it make more sense to classify different conformations first, or does it not matter if the consensus resolution is high enough?

Right now, my plan is:

1. Non-uniform refinement, with per-particle scale optimization (may or may not improve the map)
2. Local refinement (not recommended to optimize per-particle scales in local refinement)
3. Local CTF refinement
4. Global CTF refinement
5. Reference-based motion correction

Any insights on best practices for this workflow would be really helpful. Thank you!

Cheers,
cbeck

[kstachowski]

Hi @cbeck,

Thanks for your questions. Below are some of recommendations based on our experiences, so “your mileage may vary” as there are often many exceptions that will be target or dataset specific.

For your questions about when to do the above items:

• Local CTF Refinement ~4Å. For smaller particles, try limiting the defocus range to 500 to avoid deviating too much but at 3Å the default is usually good
• Global CTF Refinement (Tilt & Trefoil): ~4Å
• Global CTF Refinement (Spherical Aberration & TetraFoil) ~3-3.5Å
• Global CTF Refinement (Aniso mag): ~3-4Å
• RBMC: ~5-6Å

We would recommend a workflow like the following:

1. NU-ref +/- PPS
2. Take the best result and perform Local CTF Refinement and Global CTF refinement (all options enabled based on resolution of volumes) as separate, independent jobs (ie not within NU-ref).
3. Examine results of the CTF Refinement jobs and if changes in defoci (local CTF ref) or plots from Global CTF look reasonable (strong red and blue features with fits that match the data and low residuals) then re-run Global CTF with just the parameters that adequately correct for aberrations.
4. Run homogeneous reconstruct on all three particle sets (before, after local, after global) to check if things got better or worse.
5. If any of the above are better, add them those CTF refinements to any further NU-ref or homogeneous refinement jobs.
6. RBMC
7. Once everything is corrected for (CTF refinements and RBMC), proceed with local refinements.

If resolution improves with RBMC, it is sometimes worth assessing if higher order CTF aberrations (e.g. tetrafoil) can now be fit with more confidence. So, it might be advantageous to run another Global CTF refinement with all options enabled to see if there is any improvement.

Best,
Kye

[cbeck]

Thank you so much for this detailed response! I definitely understand that these estimates might not apply to every dataset, but it’s good to have these rough estimates as a starting point. The recommendation of 5-6 A resolution for RBMC is particularly helpful, since I was under the impression that you would want a resolution better than 3 A. Now, I know I can incorporate RBMC earlier into my workflow as soon as I get a consensus with reasonable resolution.

Would it be possible to include this information and the suggested workflow in the tutorial documentation for CTF refinement? It would be helpful for beginner to intermediate users like myself.

But before I finish off this discussion, could I ask for clarification on step 3 in you workflow? You mention re-running Global CTF with “just the parameters that adequately correct for aberrations.” Does that mean we should test each parameter independently first, then do one final Global CTF using every parameter that shows aberration in that first plot output by the first round of Global CTF jobs? I’m referring to the plot described by this section of the tutorial:

For each order of aberration (odd and even), three plots are made. The first shows the phase error data that is measured from all the particles in aggregate.

Cheers,
cbeck

[kstachowski]

@cbeck,

Now, I know I can incorporate RBMC earlier into my workflow as soon as I get a consensus with reasonable resolution.

RBMC takes a lot of computational time and will takeup additional storage space, so just ensure that you have as clean of a particle stack as possible. Having a clean particle stack also helps to better estimate hyperparameters for RBMC.

Does that mean we should test each parameter independently first, then do one final Global CTF using every parameter that shows aberration in that first plot output by the first round of Global CTF jobs?

You can test them all in the same job. All terms are fit together at the same time and if they are not enabled, the fitted values wont be retained and the plots will not be made. So you can run a job with all enabled to see which ones actually need to be fit. This will tell you which aberrations need corrected. Then enable those options in future global refinements.