Sub-tomogram averaging¶

Traditional sub-tomogram alignment-through-classification is a powerful strategy for de novo structure determination. It involves iterative 3D classification, alignment, and averaging of sub-volumes as described in Bartesaghi et al., 2008. Initially, homogeneous particle groups are identified through 3D classification and subsequently averaged in 3D. The resulting class averages are then aligned to one another and combined into high signal-to-noise (SNR) references, which can be used to align individual sub-volumes. The resulting 3D models can then serve as references for high-resolution refinements using 2D projections.

Requirements¶

• An existing Particle picking block with selected particles.

Modes of operation¶

The sub-tomogram averaging functionality in nextPYP is provided by the Sub-tomogram averaging block, which supports four primary modes of operation:

1. Mode 0 - Global averaging and iterative centering. Computes a global average of all sub-volumes, which can then be used (optionally) as a reference to iteratively center all sub-volumes using translation-only alignment. To enhance accuracy and reduce model bias, a radially symmetrized global average can be used as the reference.

2. Mode 1 - 3D classification. Based on the most recent set of alignments from the previous mode, sub-volumes are clustered into discrete classes, and class averages are computed.

3. Mode 2 - Class average alignment. Class averages are aligned to each other using a user-specified reference class. The user also selects which classes to retain. After alignment, the selected classes are averaged to produce a new reference volume.

4. Mode 3 - Sub-volume alignment to reference. Individual sub-tomograms are aligned to the reference generated in the previous step. Rotational alignment can either be global (searching the entire SO(3) space) or restricted to in-plane rotations around a pre-determined normal direction. When possible, restricting the rotation space often results in more accurate alignments.

Masks and filters for alignment and classification¶

For all modes, you can configure masking and filtering settings:

• Masking: Specify a radius in the x, y, and z directions and the apodization width, all in binned pixels.

• Filtering: Set low-pass and high-pass filter cutoffs and decay parameters, expressed as fractions of the Nyquist frequency. For example, 0.05,0.01 sets the cutoff to 0.05``and the decay to ``0.01 (0 being the DC-component and 1 being Nyquist).

Step 0. Preparation¶

1. To access sub-volumes for averaging, generate them via the Particle picking block. In the Particle extraction tab, set the Sub-volume export format to 3davg, and define the desired Sub-tomogram size (voxels) and Sub-tomogram binning.

2. Click Save, Run, and Start Run for 1 block.

Step 1. Global average and centering¶

After generating the sub-volumes, crate and configure the Sub-tomogram averaging block to run Mode 0:

1. Click on Particles (output of the Particle picking block), then choose Sub-tomogram averaging. This will create a new block and show the form to enter parameters.

2. Under Alignments from sub-volume averaging (*.txt), navigate to the frealign directory from the upstream block and select the *_original_volumes.txt file.

3. Choose mode 0 - global average and centering as the Refinement mode.

4. To use a radially symmetrized average for centering, enable Rotational symmetry, set the number of centering iterations, and adjust any masking or filtering settings.

5. Click Save, Run, and Start Run for 1 block.

6. Review results in the Global averages tab within the Sub-tomogram averaging block.

Step 2. 3D classification¶

To perform 3D classification (Mode 1):

1. Return to the project page and select Edit from the block menu.

2. Choose mode 1 - classification as the Refinement mode.

3. Set the number of desired classes and configure masking or filtering as needed.

4. Click Save, Run, and Start Run for 1 block.

5. View results in the Classes tab of the Sub-tomogram averaging block.

Step 3. Selection and alignment of class averages¶

To align selected class averages to a reference (Mode 2):

1. Return to the project page and select Edit from the block menu.

2. Choose mode 2 - alignment of averages as the Refinement mode.

3. Specify the class selection, listing the reference class first (e.g., 5,1,3,4 aligns classes 1, 3, and 4 to class 5).

4. Set masking and filtering options as needed.

5. Click Save, Run, and Start Run for 1 block.

6. Review aligned classes in the Classes (aligned) tab of the Sub-tomogram averaging block.

Step 4. Alignment of sub-tomograms to reference¶

To align all sub-volumes to the generated reference (Mode 3):

1. Return to the project page and select Edit from the block menu.

2. Choose mode 3 - alignment to reference as the Refinement mode.

3. Set the parameters for rotational and translational search, along with masking and filtering options.

4. Click Save, Run, and Start Run for 1 block.

Step 5. Iterative refinement¶

1. Return to the project page and select Edit from the block menu.

2. Increase the Iteration number to 2 and repeat steps 2-4 (in that order) to iteratively refine your model.

A typical workflow might look like:

• Iteration 1:

  • Steps 1-4 (modes 0-3)

• Iteration 2:

  • Steps 2-4 (modes 1-3)

• Iteration 3:

  • Steps 2-4 (modes 1-3)

• …continue as needed

While this process can be automated, we recommend executing each step manually to improve the quality and accuracy of the resulting 3D reconstructions.