celullarEnvTutorial.md

---

title: "Subtomogram averaging in a celullar environment" author: [Scipion Team] date: "2025-3-17" subject: "Markdown" keywords: [Markdown, Example] subtitle: "From movies to averages" lang: "en"

titlepage: true titlepage-text-color: "7137C8" titlepage-rule-color: "7137C8" titlepage-rule-height: 2 titlepage-logo: "logo.pdf" logo-width: 30mm toc: true toc-own-page: true ...

The dataset

It only contains two tilt series. This workflow is ideal to be executed in a normal laptop with GPU. The data set can be downloaded using the next command line

scipion3 testdata --download chlamy_data_tutorial

Alternatively, the dataset can be found in the next link https://scipion.cnb.csic.es/downloads/scipion/data/tests/chlamy_data_tutorial/

Import tilt series movies

Reference: J. Jimenez de la Morena et.at. 2022

Plugin: scipion-em-tomo

In this step the acquired data will be imported into the Scipion framework. The protocol tomo - import tilt series movies allows to import different kinds of raw data. In this tutorial the raw data is a set of .mrc and .mdoc files.

Note: It is mandatory that the mdoc files will be located in the same folder of the mrc files.

The .mrc files contain the acquired images, while the .mdoc files store detailed acquisition information such as sampling rate, dose per frame, and tilt axis angle. The protocol will parse the .mdoc files and store this information in the Scipion database.

Protocol parameters: Microscope voltage, spherical aberration, amplitude contrast, magnification, pixel size, tilt axis angle or dose are empty. An empty entry means that the parameter will be read from the mdoc. However, if the user introduce a parameter, Scipion will take the introduced value from the user instead of the mdoc parameter. This allows to correct possible errors in the mdoc file.

The used parameters are shown in the Figure. The critical ones are:

• Files directory: Set the path of the movies. Note that the mdoc files must be in the same folder.
• Pattern: This tutorial uses mdoc. Set as *.mdoc to import all mdoc files.
• Tomo5 mdoc: Set as Yes.
• Microscope Voltage: 300kV
• Spherical Aberration: 2.7 mm
• Amplitude contrast: 0.1
• Pixel size: 1.91.
• Tilt axis angle (deg): -95.0. By setting this parameter the pixel size from the mdoc will be overwritten.
• Dose (e/A^2): Initial dose 0.0, dose per tilt - leave empty. This value will be read from the mdoc file.
• Gain image: Set here the path to the gain image.

Movie alignment and CTF estimation with Warp

Reference: D. Tegunov et. al. 2021

Plugin: scipion-em-warp

Once the tilt series movies were imported the acquired frames at each tilt angle will be aligned to obtain tilt series. The protocol warp - tilt-series motion and ctf estimation will find and correct the relative movement between the frames. This protocol also allows to estimate the CTF, it means the defocus of ech tilt image.

The most important parameters are:

• Input movies: The imported set of tilt series movies from the previous step.
• Binning factor: 1.0
• Resolution to fit: Min 500.0 Max 10.0 A
• B-factor: -500.0
• Motion model grid: Leave empty
• Do even and odd: Yes
• Transpose gain reference: No swap
• Flip gain reference: No flip
• EER fraction: 8
• EER group exposure: Leave empty (No matter for this dataset)
• Estimate CTF: Yes.
• Windows: 512
• Resolution: (A) Min 30, Max 8.0
• Defocus search range (um): Min 0.5, Max 8.0
• Defocus model grid: X=1, Y=1, Temporal = 8
• Fit phase: No
• Use the movie average: No

The output of the protocol will be a set of tilt series and a set of CTFs. It is possible to visualize these set with the TomoViewer and CTFtomoViewer.

Excluding views and CTFs

Excluding views

Some views can be excluded with the TomoViewer. The main reasons to exclude a tilt image are: Blck image, close to the carbon border, presence of contamination, or drift. To exclude a view just select the corresponding tilt image and press the space. Alternatively, it can be marked by clicking on the exclude box. The excluded views will be highlighted in red. Finally, it is neccesary to generate a new set of Tilt Series by clicking on the botton Generate subsets.

For this dataset the low and high tilts will be removed. The first 7 negative tilts and the last 4 (positive tilts).

Excluding CTFs

There is no need to exclude CTFs for this dataset, because all views with a wrong defocus were excluded as views, however, the reader can do it if he wishes.

X-ray eraser

Reference: J.R. Kremer 1996

Plugin: scipion-em-imod

The interaction of electrons with the sample can generate X-rays. They can be detected by the camera, and identified in the images as very bright pixels. Therefore, the X-ray peaks are an unwanted effect that should be corrected. The protocol imod - Xray eraser allows to remove these bright points. The input will be a tilt series (output of the movie alignment). This protocol also can be executed with default parameters.

The ouput of this protocol will be a set of Tilt Series that looks almost identical to the input tilt series.

Tilt series alignment

There are many methods to align tilt series in the ScipionTomo framework, as they are:

• IMOD
• Aretomo
• EmanTomo

In this tutorial aretomo - tilt series align and reconstruct will be used. The user can play with imod - patch tracking to try with a different alignment.

Aretomo - tilt series align and reconstruct

Reference: S. Zheng 2022

Plugin: scipion-em-aretomo

To reconstruct the tomogram from the tilt series the protocol aretomo - tilt series align and reconstruct. This protocol performs the alignment and reconstruction at once. The result will be a set of aligned tilt series and the reconstructed tomogram. Also the CTF estimation is possible if the user select the CTF estimation option. It offers two different reconstruction algorithms: Weighted Back projection (WBP) and Simultaneous algebraic reconstruction technique (SART).

Tip: WBP is faster than the SART method, but SART provides higher contrast. To visualize cellular enviroments SART is recommended, or to pick subtomogram with a template matching approach.

The used parameters will be

• Input Tilt Series: This set will be the imported tilt series
• Skip alignment: No
• Reconstruct tomograms: Yes
• Reconstruct odd even: Yes
• Binning: 4. The tomogram will be reconstructed at bin 4
• Volume height for alignment: 1000
• Tomogram thickness: 2048
• Refine tilt angles?: Measure and correct
• Refine tilt axis angle? Refine and use the refined value for the entire tilt series
• Do dose weighting?: Yes
• Reconstruction method: WBP
• Dark tolerance: 0.7

The input of the aretomo will be the imported tilt series. To reduce the computational burden, the WBP algorithm will be chosen, and the tomomgrams will be reconstructed at binning 4.

The output can be visualized by clicking on Analyze results or alternatively by choosing the visualization tool by right-clicking on the output in the Summary box. To enhance the visualization an average of 10 slices is shown in the figure.

Tip: It can be observed from the side views of the tomogram, that the tomogram thickness is much thinner than the thickness of the tomogram. For this sample, the tomogram thickness should be around 250 px. We have aligned with Aretomo and also reconstructed to measure the thickness. Later we will reconstruct (in this tutorial with imod) with a thickness of 350px to work with a more approxiate tomogram thickness, it is recomendable to leave an extra space 50px in the top and in the botton.

CTF correction

Reference: J.R. Kremer 1996

Plugin: scipion-em-imod

Note: Scipion has a standard CTF model, when the CTF is estimated with any CTF estimator, the output is converted and stored in the Scipion standard. To correct the CTF, Scipion converts the standard into the corresponding package (in this case imod).

The CTF correction can be performed with the protocol imod - correct CTF. The input of this protocol are the tilt series with assigned alignment and a set of CTFs previously estimated. The tilt series will be the ones we assigned the alignment information. The used parameters for this protocol will be left as default parameters.

Tilt series preprocess

Reference: J.R. Kremer 1996

Plugin: scipion-em-imod

The CTF-corrected tilt series will be the input data for a later tomogram reconstruction. Up to this step we have worked with the full-size tilt series (binning 1). IF the tomograms are reconstructed at bin 1, they will be very heavy. To safe disc and enhance the SNR of the tomogram the tilt series will be binned. The protocol imod - preprocess tilt series allows to perform different preprocessing operation on the tilt series, as binning or adjusting the gray values.

Tomogram reconstruction

There are many methods to reconstruct tomograms in ScipionTomo framework, as they are:

• Tomo3d
• Imod
• AreTomo
• Emantomo

This tutorial uses IMOD.

Tomogram reconstruction with IMOD

Reference: J.R. Kremer 1996

Plugin: scipion-em-imod

To reconstruct the tomogram from the tilt series the protocol imod - reconstruct tomogram will be used. IMOD provides two different reconstruction algorithm: Weighted Back Projection (WBP) and SIRT-like filtered tomograms.

Tip: WBP is faster than the SIRT-like method, but SIRT provides higher contrast. To visualize cellular enviroments SIRT is recommended, or to pick subtomograms with a template matching approach.

The input of the reconstruction will be the binned tilt series. The Tomogram Thickness was set to 350 voxels.

The protocol can be executed with default parameters:

• Tilt series: The binned tilt series
• Tomogram Thickness: 350
• Tomogram width (voxels): 0
• Tomogram Shift: In X 0.0 in Z 0.0
• Offset (deg) of the: Tilt angles 0.0 Tilt axis 0.0
• Super-sampling factor: 2
• Iterations of a SIRT-like equivalent filter: 10
• Cutoff linear region: 0.35
• Radial fall-off: 0.035

The output can be visualized by clicking on Analyze results or alternatively by choosing the visualization tool by right-clicking on the output in the Summary box.

Denoising with cryoCARE

Reference: T.O. Bucholtz 2019

Plugin: scipion-em-cryocare

The denoising of the tomogram will be carried out with cryoCARE. This method makes use of two half tomograms, odd and even tomograms. These two tomograms should be reconstructed with with same alignment information, but with different movie frames. During the movie alignment we obtain three tilt series: The full tilt series, the odd tilt series and the even tilt series. The last two are the result of averaging the odd and even movies frames respectively. During the tilt series alignment, the full tilt series was used. Therefore, 3 tomograms were reconstructed: the full tomogram, the odd tomogram and the even tomogram. Cryocare will use the odd and even tomograms.

Cryocare workflow is split in two steps: training and prediction.

cryoCARE training

cyocARE uses a Noise2Noise model to perform the denoising. To do that,

• Are odd-even associated to the tomograms: Yes
• Tomograms: The reconstructed ones with imod
• Side lenght for training volumes: 32
• Number of training pairs to extract per tomogram?: 700
• Training epochs: 70
• Steps per epoch: 50
• Batch size: 64
• Convolution kernel size: 3
• U-net depth: 3

The output of this protocol is a new object, a cryoCARE model, that can only be used by cryoCARE. This model contains the weights of the neural network to perform the denoising in the predicition step.

Warning: The cryoCARE model is an object that cannot be visualized.

cryoCARE prediction

Once the cryoCARE model has been trained, it can be applied to the reconstructed tomograms. The parameter are:

• Tomograms: The reconstructed tomograms with imod
• cryoCARE model: The previous trained cryoCARE model
• Number of tiles: 8

Tip: This protocol can be prone to run out of memory (failure), if this happens increase the number of tiles

The result of the prediction can be visualized with the tomoViewer opening the tomograms with imod-3dmod as it is shown in the figure. Note how the ribosome are clearly visible in the tomogram.

Picking

This dataset can be picked with any of the next software packages

• Sphire - cryolo
• Emantomo - template matching
• Gapstop - template matching
• Deepfinder

For this tutorial, emantomo - template matching will be used. This picking requires a refence, for that reason a ribosome reconstruction from EMDB will be download

Import a reference

A reference can be imported into Scipion from a local file or from an EMDB entry. In this case we will download a Ribosome map from EMDB, the entry EMD-50827

The imported volume can be visualized in slices or in chimera

Resize the reference

To perform the template matching, the reference and the tomogram must be at the same pixel size. For this reason a resize step is needed. The reconstructed and denoised tomograms present a pixel size of A/px, . In addition, the imported volume has a boxsize much larger than the protein diameter. In order to speed up the template matching, a croping step will be also carried out. The protocol xmipp - crop and resize will be used with the next parameters:

• Input Volumes: The imported one
• Resize volumes?: Yes
• Resize option: Sampling Rate
• Sampling rate: 7.64A/px
• Crop volumes?: Yes
• Window operation: Window
• Window size: 54

Emantomo template matching

Reference: M. Chen et.al 2019

Plugin: scipion-em-emantomo

It is neccesary to identify the proteins in the tomograms. The protocol emantomo - template matching will be used. for these tomograms the next parameters will be used:

• Input tomograms: The reconstructed ones
• Template: The imported and resized volume
• Maximum no. particles picked among the tomogams: 800
• Distance threshold: -1
• Template matching threshold (n sigma): 6.0
• Symmetry of the reference: c1
• Remove particles on the edge: No
• Remove particles near gold fiducials: No

The results can be opened with EmanTomo, Napari or Dynamo viewers. With a high number of particles we have observed some crashes in Eman viewer. That should not happen in this tutorial

Subtomogram Averaging with RelionTomo

Reference: A. Burt 2024

Plugin: scipion-em-reliontomo

Extract pseudo-subtomograms at bin 8

Pseudo-subtomograms are 2D cropped images or 3D-reconstructed subtomograms, premultiplied by the CTF. They do not represent physical objects. The premultiplication by the CTF, allows to avoid aliasing and speed up the computational times. The protocol reliontomo - extract subtomos can be used to extract the pseudo-subtomograms. This protocol allows to extract them as 2D or 3D. In this tutorial both kinds of psuedo subtomogram will be extracted. The 3D pseudo-subtomogram will be used for generating an initial volume, and the 2D pseudo-sutomograms will be used in the refinement step.

To extract the pseudo subtomograms the next input data will be required:

• Tilt series: They must contain the alignment information.
• CTF estimation: From the Warp estimation with the excluded views
• Coordinates: They are the picked coordinates.
• Binning factor: 8 This is the scaling factor in relation to the input tilt series
• Box size (px): 64. This box size will be used to correct the CTF in the cropped particles from the tilt series
• Croppped box size (px): 32. This will be the size of the pseudo-subtomograms and therefore of the reconstructed map.
• Write output as 2D stacks: No.

Tip: The 3D pseudo-subtomograms work better for obtaining a 3D initial model than the 2D pseudo-subtomograms.

In he extracted pseudo subtomograms can barely be visualized the ribosomes. However, the output of the protocol also provides a landmarkmodel, with the positions of the picked particles on the tilt series.

Initial model

The initial model can be estimated with the protocol reliontomo - 3D initial model. The input will be the extracted 3D pseudo-subtomogram at bin 10 from the previous step.

• Number of VDAM mini-batches: 40. This is the number of iterations to be carried out.
• Regularization parameter: 4.0 It goes from 0 to 4. Values close to 4 put more strenght on the data.
• Circular Mask diameter: 330 A. A good value is to set the protein diameter
• Symmetry group: C1. In this case the protein has no symmetry. For initial volumes a C1 symmetry is a good practice.
• Prior width on tilt angle: -1. It defines the prior on the tilt to be estimated. -1 means no priors

3D Auto-refine at bin 8

Using the initial model, it is possible to refine it to enhance the map quality pushing the resolution. Despite at binnin 8 the initial volume presents enought quality, we will enhance a litle bit more the map reaching Nyquist resolution, and then in a later step extract the pseudo-subtomogram at a smaler pixel size. To refine the model, the protocol reliontomo - 3D auto-refine can be used. The input will be the extracted 3D pseudo-subtomogram at bin 8 and the estimated initial model. The refinement parameter will be.

• Pseudo-subtomograms: The extracted 3D-pseudosubtomograms. Despite for refining the 2D-psedosubtomograms are recommended, at this so low resolution the 3D will be used.
• Reference volume: The obtained initial volume
• Is initial 3D map on absolute greyscale?1: Yes
• Resize references if needed?: Yes
• Initial low-pass filter (A): 40A
• Symmetry group: C1
• Do CTF-correction?: Yes
• Ignore CTF until first peak?: No
• Circular Mask diameter: 350A.
• Mask particles with zeros: Yes
• Use blush regularization: No
• Initial angular sampluing interval: 7.5 deg
• Initial offset range (px): 5 px
• Initial offset step (px): 1 px
• Local searches from auto-sampling: 1.8 deg
• Symmetry to be relaxed: Leave empty
• Use finer angulat sampling faster: No
• Prior width on tilt angle: -1

The result of this protocol should be similar to the one shown in the Figure.

Extract pseudo-subtomograms at bin 4

This steps shows how to reduce the binning keeping the alignment of already refined pseudo-subtomograms. The protocol reliontomo - extract subtomos allows this task. The parameters:

• Coordinates: They will be the refined pseudo-subtomograms from the 'reliontomo - 3d auto-refine`.
• CTF: The estimated with Warp with excluded views.
• Tilt series: The aligned ones with dose and excluded views.
• Binning: 4.0.
• Box size (px): 128 px. This box size will be used to correct the CTF in the cropped particles from the tilt series
• Croppped box size (px): 64 px. This will be the size of the pseudo-subtomograms and therefore of the reconstructed map.
• Maximum dose: 50 e/A^2.
• Write output as 2D stacks: No

Reconstruct particle at bin 4

In this step the refined pseudo-subtomograms from the previous autorefine are used to reconstruct the protein, but keeping their angular assignment. This is only a reconstruction step. The protocol reliontomo - reconstruct particle

• Coordinates/Pseudo-subtomograms: They will be the refined pseudo-subtomograms from the 'reliontomo - 3d auto-refine`.
• Binning: 4.0.
• Box size (px): 128 px. This box size will be used to correct the CTF in the cropped particles from the tilt series
• Croppped box size (px): 64 px. This will be the size of the pseudo-subtomograms and therefore of the reconstructed map.
• Symmetry group: C1.
• Apply Wiener filter with SNR: 0.0

The reconstructed protein can be visualized with Scipion (to see the slices) or with Chimera (to see the 3D map). As it can be observed in the figure the map quality enhanced in comparison to the reconstruction at bin 8.

Refine volume at bin 4

Now the obtained reconstruction will be refined with the aim of pushing the resolution of the reconstructed map. The protocol reliontomo - extract subtomos will be used with the next parameters

• Pseudo-subtomograms: The 3D extracted ones at bin 4
• Reference Volume: The reconstructed volume at bin 4 from the previous step
• Reference Mask: Leave empty, we lack of resolution for it
• Is initial 3D map on absolute greyscale?1: Yes
• Resize references if needed?: Yes
• Initial low-pass filter (A): 17A
• Symmetry group: C1
• Do CTF-correction?: Yes
• Ignore CTF until first peak?: No
• Circular Mask diameter: 320A.
• Mask particles with zeros: Yes
• Use blush regularization: No
• Initial angular sampluing interval: 7.5 deg
• Initial offset range (px): 5 px
• Initial offset step (px): 1 px
• Local searches from auto-sampling: 1.8 deg
• Symmetry to be relaxed: Leave empty
• Use finer angulat sampling faster: No
• Prior width on tilt angle: -1 Meaning no prior

Contact us

We want to hear from you! Any comment, question, or complaints regarding this tutorial, the use of Scipion can be sent to the next email: scipion@cnb.csic.es.

We also have a discord server where a cryoEM/ET community is active and in touch daily, please send us an email to get an invitation.

References

• JM De la Rosa-Trevín, A Quintana, L Del Cano, et al. Scipion: A software framework toward integration, reproducibility and validation in 3D electron microscopy, Journal of Structural Biology, 195,1, 93-99 (2016).
• A. Burt, C.K. Cassidy, P. Ames, P. et al. Complete structure of the chemosensory array core signalling unit in an E. coli minicell strain. Nat Commun 11, 743 (2020).
• J.R. Kremer, D.N. Mastronarde, J.R McIntosh, Computer Visualization of Three-Dimensional Image Data Using IMOD, Journal of Structural Biology, 116, 1, 71-76 (1996)
• D.N. Mastronarde, S.R. Held, Automated tilt series alignment and tomographic reconstruction in IMOD, Journal of Structural Biology, 197, 2, 102-113 (2017)
• JI Agulleiro, JJ Fernandez. Fast tomographic reconstruction on multicore computers. Bioinformatics 27:582-583, (2011).
• JI Agulleiro, JJ Fernandez. Tomo3D 2.0--exploitation of advanced vector extensions (AVX) for 3D reconstruction. Journal of Structural Biology 189:147-152, (2015).
• A. Rohou, N. Grigorieff, CTFFIND4: Fast and accurate defocus estimation from electron micrographs, Journal of Structural Biology, 192, 2, (2015)
• M. Chen, J.M. Bell, X. Shi, X. et al. A complete data processing workflow for cryo-ET and subtomogram averaging. Nat Methods 16, 1161–1168 (2019)
• Q. Xiong, M.K. Morphew, C.L. Schwartz, CTF Determination and Correction for Low Dose Tomographic Tilt Series, Journal of Structural Biology, 168(3) 378–387 (2009).