In the previous post, we covered how to install RELION 5.0 and set up the Conda environment.
In this article, we will follow the official RELION tutorial step by step, reproducing the workflow through the GUI with the same input parameters and job types.

The dataset used here is the official RELION tutorial dataset (relion30_tutorial_data.tar),
which contains β-galactosidase images collected on a JEOL Cryo-ARM (200 kV) microscope.

1. Download the Tutorial Dataset

Download and extract the tutorial data:

wget ftp://ftp.mrc-lmb.cam.ac.uk/pub/scheres/relion30_tutorial_data.tar
tar -xf relion30_tutorial_data.tar

This creates a folder relion30_tutorial_data/Movies/ containing the movie files.

2. Project Setup and Import

Launch RELION in the desired working directory. When prompted that no project exists, confirm to create a new one.

  • Job Type: Import
  • Movies / Micrographs (I/O tab): /Movies/*.tiff
  • Import raw micrographs? Yes
  • MTF of the detector: mtf_k2_200kV.star
  • Pixel size (Å): 0.885
  • Voltage (kV): 200
  • Spherical aberration: 1.4
  • Amplitude contrast: 0.1

After running the job, an Import/job001/ directory is created and micrographs are registered in the project.

3. Motion Correction (Frame Alignment)

Although the tutorial dataset already includes motion-corrected micrographs, you can practice setting up the Motion Correction job as follows:

  • Job Type: Motion correction (MotionCor2)
  • Input (I/O tab): Import/job001/movies.star
  • Dose per frame (e⁻/Ų): 1.277
  • EER fractionation: 32
  • Write output in float16: Yes
  • Do dose-weighting: Yes
  • Save sum of power spectra: Yes
  • Sum of power spectra every e/Ų: 4

In the Motion tab:

  • Gain-reference image: Movies/gain.mrc

The tutorial recommends running with MPI = 1, threads = 12. For a workstation with 24 CPU cores, you can use MPI = 3 and threads = 4–6. The job outputs corrected averages and .star files under MotionCorr/job002/.

4. CTF Estimation

Estimate the CTF parameters from the motion-corrected micrographs.

  • Job Type: CTF estimation (CTFFIND4)
  • Input (I/O tab): MotionCorr/job002/corrected_micrographs.star
  • Maximum resolution (Å): 3
  • Defocus search range (Å): 5000–20000

Results are stored in CtfFind/job003/, including .star files and per-micrograph CTF information. Use the Display panel to inspect Thon rings and defocus fits.

5. Auto-picking

Perform automatic particle picking using the Laplacian of Gaussian (LoG) method.

  • Job Type: Auto-picking (LoG)
  • Input micrographs (I/O tab): CtfFind/job003/micrographs_ctf.star
  • Min. diameter for LoG filter (Å): 150
  • Max. diameter for LoG filter (Å): 180

For efficiency, pick only a subset of micrographs using “Subset selection”. These coordinates can be used later as templates for Topaz-based auto-picking.

6. Particle Extraction

Extract particles based on the detected coordinates.

  • Job Type: Particle extraction
  • Input micrographs: CtfFind/job003/micrographs_ctf.star
  • Input coordinates: AutoPick/job004/particles.star
  • Box size: 256 pixels
  • Rescaling / Downsampling: 128 pixels

The extracted particles.star file is saved in Extract/job005/ and will be used for 2D classification.

7. 2D Classification

Perform 2D classification to assess particle quality and remove outliers.

  • Job Type: 2D classification
  • Input particles: Extract/job005/particles.star
  • Number of classes (K): 50
  • Regularisation parameter (T): 2
  • Number of iterations: 25
  • Initial reference: None / default

Results are stored in Class2D/job006/. Use the Display panel to inspect class averages, and select good classes via “Subset selection”. These refined particles can then be reused for Topaz training or higher-quality picking.

8. Initial 3D Model (VDAM-based Ab-initio)

Generate an initial 3D model using selected particles.

  • Job Type: 3D initial model
  • Input particles: selected .star file
  • Number of classes: 1
  • Symmetry: C1
  • Mask diameter (Å): auto or specified
  • Low-pass filter (Å): 30

This creates the initial 3D volume under InitialModel/job007/.

9. 3D Refinement

Refine the structure using the full dataset and the initial model.

  • Job Type: 3D auto-refinement
  • Input: selected .star file
  • Mask: auto-generated or custom
  • Symmetry: D2 (the tutorial dataset has D2 symmetry; use C1 otherwise)
  • Initial low-pass filter (Å): 50

The final refined map run_class001.mrc will appear in Refine3D/job008/.

10. Post-processing and Analysis

  • Map sharpening / B-factor correction to enhance contrast
  • Local resolution estimation (LocalResolution job)
  • CTF refinement / higher-order aberration correction
  • Map visualization: ChimeraX or UCSF Chimera
chimerax Refine3D/job008/run_class001.mrc

Inspect the FSC curve in postprocess.star (0.143 cutoff criterion). Class subsets can also be managed via the GUI’s Display → Subset selection tool.

The official RELION tutorial also includes VDAM-based modeling, Blush Regularisation, CTF refinement, polishing, and higher-order aberration corrections for advanced refinement.

In general, a standard SPA workflow proceeds as follows:

  1. Import micrographs
  2. Motion correction
  3. CTF estimation
  4. Particle extraction
  5. 2D classification
  6. Particle selection
  7. Repeat 4–6 until good particles are obtained
  8. Initial model generation
  9. 3D classification and refinement
  10. Mask creation
  11. Post-processing (sharpening, CTF refinement, polishing)

After these steps, higher resolution can be achieved through iterative refinement. For reference, always consult the official RELION tutorial for detailed updates.

In the next post, we will explore CryoSPARC, focusing on its workflow and comparison with RELION.