Skip to main content

EMPIAR-10985 tutorial

Obtaining a subnanometer subtomogram average in 9h

  • Time requirement: 9 h compute time | 20 min manual time
  • Goal: Determine a 9 angstrom structure of the bacterial 70S ribosome by subtomogram averaging.

This tutorial will walk you through the analysis of the EMPIAR dataset 10985. The linked dataset contains 20 tilt-series of the purified bacterial 70S ribosome.

Tilt-series alignment & tomogram reconstruction​

The tilt-series job on CryoCloud runs all steps from motion correction to tomogram reconstruction, including tilt-series alignment and CTF estimation. Currently, all mdoc files must be present in the same folder as the movies.

To run the job, specify the following parameters.

Input:

  • Micrographs directory: datasets/199/

Motion correction:

  • Dose per frame: 0.5

Tilt-series alignment:

  • Voltage: 300
  • Pixel size: 1.07
  • Axis rotation: 5

Tomogram reconstruction:

  • Reconstruct tomograms: Yes
  • Binning factor: 8
  • Thickness: 1600

To inspect the resulting tomograms, download them from the results folder and visualize them in a 3D viewer of your choice. You can find the tomograms in the tomogram subfolders of the results section with the suffix _rec.mrc.

3D Template matching​

Before we can continue with the subtomgoram averaging of ribosomes, we need to localize them in the tomograms first. One effective technique for the automated localization of particles is 3D template matching. It requires a template of your target molecule, which will then be used to calculate the cross correlation between thousands of different orientations of your template and the tomogram at each voxel. To perform this task, we will use the pytom tool for template matching (Chaillet et al. 2023).

Please download the following map and mask of the 70S ribosome and re-upload them to your project archive:

Once you uploaded the files to the archive, create a Template Matching job in the Template Matching column using the following parameters: Input:

  • Tomograms STAR file: jobs/TiltSeries/<N>/tomograms.star

Template matching:

  • Template: archive/<downloaded-template>.mrc
  • Mask: archive/<downloaded-mask>.mrc
  • Tilt angles: -60, 60
  • Angular search: 12.8

Extraction:

  • Particle radius: 18
  • Binning factor: 8
  • Number of particles: 400

The output from this job are volumes with the cross correlation scores and angles for each voxel, as well as a coordinate file that you can use as input for the Extract Subtomo job. To inspect the quality of the picks, download and visualize the volumes with the suffix _scores.mrc in the 3D viewer of your choice, and overlap them with your tomogram.

Subtomogram Extraction​

Next, we need to extract the ribosomes at the coordinates identified during the template matching. Create and run an Extract Subtomo job from the Template Matching column using the following parameters:

  • Unbinned box size: 480
  • Cropped box size: 80
  • Binning factor: 6
  • Extract in float 16?: Yes

This will extract 8,000 subtomograms as well as a particles.star file which you can use in the next 3D classification job.

3D classification​

The extracted subtomomgrams will contain some contaminations as well as large ribosomal subunits. The 3D classification job will allow us to obtain a clean set of fully assembled 70S ribosomes.

Download the following reference for your 3D classification job and add it to your project Archive to be used as a reference:

Set up and run a 3D classification job using the following parameters:

  • Input images STAR file: jobs/ExtractSubtomo/<N>/particles.star
  • Reference map: archive/<downloaded-reference>.mrc
  • Initial low pass filter: 60
  • Number of classes: 4
  • Mask Diameter: 300
  • Trust reference pixel size?: Yes

You will obtain 4 classes estimated at 13 angstrom resolution each.

3D class selection​

After the 3D classification, open a select job and specify the model.star file from the last iteration of the 3D classification job. Click Load Selector, and select the classes representing fully assembled ribosomes. You should obtain about 3,300 subtomograms.

Re-extraction of subtomograms​

To obtain a higher resolution average we need to re-extract the subtomograms at a higher pixel-size. Set up an Extract Subtomo job with the following parameters and click run:

Input:

  • Tomograms STAR file: jobs/TiltSeries/<N>/tomograms.star
  • Coordinates STAR file: jobs/Select/<N>/classes_selected.star

Extraction parameters:

  • Unbinned box size: 480
  • Cropped box size: 160
  • Binning factor: 3
  • Extract in float 16?: Yes

3D Refinement​

After re-extracting the subtomograms at a smaller pixel, we can now setup a 3D refinement job to obtain a subtomogram average at a higher resolution.

Input:

  • Input images STAR file: jobs/ExtractSubtomo/<N>/particles.star
  • Reference map: jobs/Rescale/<N>/rescaled_volume.mrc (Select the nicest class of the last iteration from the 3D classification job)
  • Input optimization set: jobs/ExtractSubtomo/<N>/optimisation_set.star
  • Initial low pass filter: 40
  • Mask Diameter: 300
  • Trust reference pixel size?: Yes

The 3D Refinement job should results in an average around 9 angstrom and finish in 6 hours. CryoCloud is currently using RELION 4 for the subtomogram averaging (Zivanov et a. 2022), and will be updated soon with the expected runtime improvements of a factor 6 (Burt et al. 2024)!

To visualize the results, download the map and open it locally in your preferred 3D map viewer.