Tomo refinement 1: CTF refinement

In this step we will show how to apply the refinement of the different parameters related to the CTF estimation such as tilt series projection defocus and contrast scale. For a full description of the arguments, check the relion_tomo_refine_ctf program.

If you are running this job without following the tutorial, please note the requirements for the reference map and FSC data as described in Reference map and FSC data.

Running the job

Since we ran both Reconstruct particle and Post-processing after the bin 1 refinement step at the end of the High-resolution 3D refinement section, we already have the required reference halfmaps (Reconstruct/job025/half<12>.mrc) and post-processing FSC data (PostProcess/job026/

In addition, we can use the optimisation set file from a previous job or, in this case, the explicit particle set from the duplicate removal step and the separate tomogram set (which would otherwise be included in the optimisation set).

Lastly, the reference mask file used in the previous 3D auto-refine at binning factor 1 is also required by the tomo refinement jobs, which in our case is mask_align.mrc.

Select the CTF refinement job-type and set on the I/O tab:

Input optimisation set::


OR: use direct entries?:


(Because the previous job run was Subset selection to remove duplicate particles, we will use the resulting particles file as the input particle set. Otherwise, we could have used the optimisation set file directly from the previous 3D auto-refine job.)

Input particle set::


Input tomogram set::


Input trajectory set:


(This is empty in the first tomo refinement cycle, unless Bayesian polishing is run first, in which case we would include the generated file, unless it already is included in the optimisation set file.)

One of the 2 reference half-maps::


Reference mask::


Input postprocess STAR::


On the Defocus tab, set:

Box size for estimation (pix):


Refine defocus?:


Defocus search range (Å):


Do defocus regularisation?:


Defocus regularsation lambda:


Refine constrast scale?:


Refine scale per frame?:


Refine scale per tomogram?:


On the Running tab, set:

Number of MPI procs::


Number of threads::


With these parameters, the job should take around 10 minutes to run.

Analysing the results

The output folder CtfRefine/job027 contains a new file with the refined parameters. To assess the result, run new Reconstruct particle and Post-processing jobs using the generated CtfRefine/job027/ file. In our workspace, we see a slight improvement in the resolution to 3.87Å.

These reference map and postprocess files will also be used as inputs for the next Bayesian polishing run.