8. Further processing
In the typical workflow we would do several further steps to maximise the quality of the map produced from your data. These include:
CTF and aberration refinement
Per-particle defocus estimation
Bayesian polishing
8.1. CTF and aberration refinement
Next, we’ll use the CTF refinement job-type to estimate the asymmetrical and symmetrical aberrations in the dataset; whether there is any anisotropic magnification; and we’ll re-estimate per-particle defocus values for the entire data set. Running this job-type can lead to further improvements in resolution at a relatively minor computational cost, but it all depends on how flat your ice was (for per-particle defocus estimates), and how well you had aligned your scope (for the aberrations). It runs from a previous 3D auto-refine job as well as a corresponding Post-processing job. Let’s start with the higher-order aberrations, to see whether this data suffered from beamtilt or trefoil (which are asymmetric aberrations), or from tetrafoil or an error in spherical aberration (which are symmetric aberrations).
Under CTF Refinement select a new RELION Ctf Refinement job:
Particles (from Refine3D):: Refine3D/job019/run_data.star
Postprocess STAR file:: Postprocess/job021/postprocess.star
Minimum resolution for fits (A):: 30
Perform CTF parameter fitting?:: Yes
Fit defocus?:: Per-particle
Fit astigmatism?:: Per-micrograph
Fit B-factor?:: No
Fit phase-shift?:: No
Estimate beamtilt?:: Yes
Also estimate trefoil?:: Yes
Estimate 4th order aberrations?:: Yes
HEADER: Running options
Number of MPI procs:: 1
Number of threads:: 24
Press RUN, the job should take ~1 minute to run. Now switch to RESULTS tab.
You’ll see that this data suffered from some beamtilt: one side of the asymmetrical aberration images is blue, whereas the other side is red. There was also a small error in the spherical aberration, as the symmetrical aberration image shows a significant, circularly symmetric difference (the image is blue at higher spatial frequencies, i.e. away from the centre of the image). Importantly, for both the asymmetric and the symmetric aberrations, the model seems to capture the aberrations well.
If the data had suffered from trefoil, then the asymmetric aberration plot would have shown 3-fold symmetric blue/red deviations. If the data had suffered from tetrafoil, then the symmetric aberration plot would have shown 4-fold symmetric blue/red deviations. Examples of those are shown in the supplement of our 2019 publication (Falcon et al) on Tau filaments from the brain of individuals with chronic traumatic encephalopathy (CTE).
8.2. 3D Refinement and PostProcessing with CTF Refined Particles
It is probably a good idea to re-run 3D auto-refine and Post-processing at this stage, so we can confirm that the new particle STAR file actually gives better results.
Let’s make life easy. In the lefthand window go to the JOBS tab and select job Refine3D/job019/. Now select the three dot menu and click Clone Job in the context menu.
This creates a new job with the same parameters as the original job. Change two of them:
Input images STAR file:: CtfRefine/job022/particles_ctf_refine.star
Reference map:: Refine3D/job019/run_class001.mrc
Press RUN. The job should take ~10 minutes to run. We’ll skip the RESULTS and go straight to the next step. In the lefthand window go to the JOBS tab again and select job PostProcess/job021/. Now, again, select the three dot menu and click Clone Job in the context menu. This will again create a new job with the same parameters as your previous PostProcess job. Change following parameter:
One of the 2 unfiltered half-maps:: Refine3D/job023/run_half1_class001_unfil.mrc
Press RUN. The job should take ~10 seconds to run on the tested system. Now go to the RESULTS tab and see how the CTF refinement has improved the reconstruction. The improvement here will be minimal, in order of 0.1 or 0.01 Angstrom.
8.3. Bayesian polishing
Relion also implements a Bayesian approach to per-particle, reference-based beam-induced motion correction. This approach aims to optimise a regularised likelihood, which allows us to associate with each hypothetical set of particle trajectories a prior likelihood that favors spatially coherent and temporally smooth motion without imposing any hard constraints. The smoothness prior term requires three parameters that describe the statistics of the observed motion. To estimate the prior that yields the best motion tracks for this particular data set, we can first run the program in ‘training mode’. Once the estimates have been obtained, one can then run the program again to fit tracks for the motion of all particles in the data set and to produce adequately weighted averages of the aligned movie frames.
8.3.1. Polishing training (Optional)
NOTE: If you’re short on time, skip this step and use the precalculated training weights.
This step calculates the optimal parameters for particle polishing. This step can take up to two hours to run.
Under Particle Polishing select a new RELION Bayesian Polishing – Training job:
Micrographs (from MotionCorr):: MotionCorr/job002/corrected_micrographs.star
Particles (from Refine3D or CtfRefine):: Refine3D/job023/run_data.star
NOTE: These particles will be polished.
Postprocess STAR file:: PostProcess/job024/postprocess.star
NOTE: The mask and FSC curve from this job will be used in the polishing procedure.
First movie frame:: 1
Last movie frame:: -1
NOTE: Some people throw away the first or last frames from their movies. Note that this is not recommended when performing Bayesian polishing in Relion. The B-factor weighting of the movie frames will automatically optimise the signal-to-noise ratio in the shiny particles, so it is best to include all movie frames.
Use this many particles:: 3500
NOTE: That’s almost all we have anyway. Note that the more particles, the more RAM this program will take. If you run out of memory, try training with fewer particles. Using much fewer than 4000 particles is not recommended.
Fraction of Fourier pixels for testing:: 0.5
Sigma for acceleration (A/dose):: 2
HEADER: Running options
Number of threads:: 24
Press RUN The job could take anywhere from 30 minutes to 2 hours to run. It run in about 1 hour on the tested system with half the threads.
8.3.2. Particle Polishing
NOTE: If you skipped the above polish training step we will perform the particles polishing step using pre-calculated parameters.
Under Particle Polishing select a new RELION Bayesian Polishing job:
Micrographs (from MotionCorr):: MotionCorr/job002/corrected_micrographs.star
Particles (from Refine3D or CtfRefine):: Refine3D/job023/run_data.star
NOTE: These particles will be polished.
Postprocess STAR file:: PostProcess/job024/postprocess.star
NOTE: The mask and FSC curve from this job will be used in the polishing procedure.
First movie frame:: 1
Last movie frame:: 20
NOTE: Here we use only the first 20 frames, to use all frames set this parameter to -1
Extraction size (pix in unbinned movie):: 360
Re-scaled size (pixels):: 256
Optimised parameter file:: Polish/job025/opt_params_all_groups.txt
NOTE: If you did run step 8.3.1, use this parameter and set `OR use your own parameters?` below to No.
OR use your own parameters?:: Yes
NOTE: If you didn’t run step 8.3.1 use this parameter.
Sigma for velocity (A/dose):: 0.438
NOTE: If you didn’t run step 8.3.1 use this parameter.
Sigma for divergence (A):: 1695
NOTE: If you didn’t run step 8.3.1 use this parameter.
Sigma for acceleration (A/dose):: 3.345
NOTE: If you didn’t run step 8.3.1 use this parameter.
HEADER: Running options
Number of MPI procs:: 1
Number of threads:: 24
Then hit RUN. The job should take about 5-10 minutes to run on the tested system.
8.4. Final processing steps
Finally run one more set of 3D refinement and post processing.
First 3D Refinement by cloning job Refine3D/job023, update the following parameter:
Input images STAR file:: Polish/jobXXX/shiny.star
NOTE: The job number will depend on whether you ran step 8.3.1 (job026) or not (job025).
Once it is finished, run Postprocessing job by cloning Postprocess/job024, update the following parameter:
One of the 2 unfiltered half-maps:: Refine3D/jobXXX run_half1_class001_unfil.mrc
Once it is finished, inspect the RESULTS tab. The resolution should have improved to something like 3.0 Angstrom.
If you’re having trouble viewing the model in Mol*, check out these solutions.
8.5. When and how to run CTF refinement and Bayesian polishing
Both Bayesian polishing and CTF refinement, which comprises per-particle defocus, magnification and higher-order aberration estimation, may improve the resolution of the reconstruction. This raises a question of which one to apply first.
In this example, we first refined the aberrations, the magnification, and then the per-particle defocus values. We then followed up with polishing, but we could have also performed the polishing before any of the CTF refinements. Both approaches benefit from higher resolution models, so an iterative procedure may be beneficial. For example, one could repeat the CTF refinement after the Bayesian polishing. In general, it is probably best to tackle the biggest problem first, and some trial and error may be necessary.
Moreover, we have seen for some cases that the training procedure of Bayesian polishing yields inconsistent results: i.e. multiple runs yield very different sigma values. However, we have also observed that often the actual sigma values used for the polishing do not matter much for the resolution of the map after re-refining the shiny particles. Therefore, and also because the training is computationally expensive, it may be just as well to run the polishing directly with the default parameters i.e. without training for your specific data set.
You should now have a beautiful reconstruction and are now ready to start model building!