Segmentation is a key step in processing cryo-electron tomography (cryoET) datasets that entails identifying specific structures of interest within the volumetric data and marking them as distinct features. This process forms the basis of many subsequent analysis steps, including particle picking for subtomogram averaging and the generation of 3D macromolecular feature maps that help visualize the ultrastructure of the sample.

Although it is a common task, the currently available software packages for tomogram segmentation often leave room for improvement in either scope, accessibility, or open availability of the source code. Some popular programs, such as Amira (Thermo Fisher Scientific), are not free to use, while other more general purpose EM data processing suites offer limited functionality in terms of visualization and user interaction. Specifically, we found that a deficit existed of software that is easy to use, competitively performing, freely available, and dedicated to segmentation of cryoET datasets for downstream processing.

In this report, we present Ais, an open-source tool that is designed to enable any cryoET user – whether experienced with software and segmentation or a novice – to quickly and accurately segment their cryoET data in a streamlined and largely automated fashion. Ais comprises a comprehensive and accessible user interface within which all steps of segmentation can be performed, including: the annotation of tomograms and compiling datasets for the training of convolutional neural networks (CNNs), training and monitoring performance of CNNs for automated segmentation, 3D visualization of segmentations, and exporting particle coordinates or meshes for use in downstream processes. To help generate accurate segmentations, the software contains a library of various neural network architectures and implements a system of configurable interactions between different models. Overall, the software thus aims to enable a streamlined workflow where users can interactively test, improve, and employ CNNs for automated segmentation. To ensure compatibility with other popular cryoET data processing suites, Ais employs file formats that are common in the field, using .mrc files for volumes, tab-separated .txt or .star files for particle datasets, and the .obj file format for exporting 3D meshes.

To demonstrate the use of Ais, we outline its use in two such tasks: first, to automate the particle picking step of a subtomogram averaging workflow, and second for the generation of rich three-dimensional visualizations of cellular architecture, with ten distinct cellular components, based on cryoET datasets acquired on cellular samples.

The first step in image segmentation using CNNs is to manually annotate a subset of the data for use as a training dataset. Ais facilitates this step by providing a simple interface for browsing data, drawing overlays, and selecting boxes to use as training data (Figure 1A, Figure 1—figure supplements 1–4). Multiple features, such as membranes, microtubules, ribosomes, and mitochondrial granules, can be segmented and edited at the same time across multiple datasets (even hundreds). These annotations are then extracted and used as ground truth labels upon which to condition multiple neural networks, each trained to segment a single feature type, with which one can automatically segment the same or any other dataset (Figure 1B). Segmentation in Ais is performed on-the-fly and can achieve interactive framerates, depending on the size of the datasets and network architectures that are used. With a little experience, users can generate a training dataset and then train, apply, and asses the quality of a model within a few minutes (Figure 1C), including on desktop or laptop Windows and Linux systems with relatively low-end GPUs (e.g. we often use an NVIDIA T1000).

Figure 1 with 4 supplements see all

Download asset Open asset

Many cryoET datasets look alike, especially for cellular samples. A model prepared by one user to segment, for example, ribosomes in a dataset with a pixel size of 10 Å, might also be adequate for another user’s ribosome segmentation at 12 Å per pixel. To facilitate this sort of reuse and sharing of models, we launched an open model repository at aiscryoet.org where users can freely upload and download successfully trained models (Figure 1D). Models that pass screening become public, are labelled with relevant metadata (Figure 1E), and can be downloaded in a format that allows for direct use in Ais. Thus, users can skip the annotation and training steps of the segmentation workflow. To kickstart the repository, all 27 models that are presented in this article have been uploaded to it.

Our software is not the first to address the challenge of segmenting cryoET datasets; established suites such as EMAN2 (Galaz-Montoya et al., 2015), MIB (Belevich et al., 2016), SuRVOS (Luengo et al., 2017), or QuPath (Bankhead et al., 2017) also provide some or most of the functionality that is available in Ais. Each comes equipped with one or various choices of neural network architectures to use, and many more designs for neural networks for semantic image segmentation can be found in the literature. Therefore, as well as creating a package geared specifically towards ease of use and fast results, we also wanted to include functionality that enables a user to quickly compare different models in order to facilitate determining which models are best suited for a particular segmentation task. The software thus includes a library of a number of well-performing architectures, including adaptations of single-model CNN architectures such as InceptionNet (Szegedy et al., 2014), ResNet (Szegedy et al., 2016), various UNets (Ronneberger et al., 2015), VGGNet (Simonyan and Zisserman, 2015), and the default architecture available in EMAN2 Galaz-Montoya et al., 2015, as well as the more complex generative-adversarial network Pix2pix (Isola et al., 2017). This library can also be extended by copying any Python file that adheres to a minimal template into the corresponding directory of the project (Appendix 1).

A library of neural network architectures supports varied applications

To illustrate how useful it can be to rapidly test various architectures before selecting one that is well suited for the segmentation of any particular feature, we used six different architectures for the segmentation of three distinct features within the same tomogram, and analyzed the results (Table 1). We used a cryoET dataset that we had previously acquired (Abendstein et al., 2023), which contained liposomes with membrane-bound Immunoglobulin G3 (IgG3) antibodies that form an elevated Fragment crystallizable (Fc) platform, prepared on a lacey carbon substrate. The features of interest for segmentation were the membranes, antibody platforms, and carbon support film (Figure 2A).

Figure 2 with 1 supplement see all

Download asset Open asset

Table 1

Architecture	Parameters	Training timea	Processing timea, b	Membranec	Carbonc	Antibodyc
EMAN2	378,081	38 s	50 ms	0.044	.072	.010
InceptionNet	550,529	462 s	295 ms	0.046	.120	.008
UNet	922,881	132 s	110 ms	0.013	.007	.001
VGGNet	1,493,059	91 s	85 ms	0.011	.125	.013
ResNet	4,887,265	1533 s	980 ms	0.047	.060	.014
Pix2pixd	29,249,409	793 s	225 ms	0.050d	.129d	.016d

After training, we applied the resulting models to a different tomogram containing the same features, that were not previously used to generate the training data (Figure 2B), so that there was no overlap between the training and testing datasets. Next, we compared the training times, relative loss values, and quality of the segmentations.

Based on the loss values (the loss is a metric of how well predictions match the ground truth labels), VGGNet was best suited for the segmentation of membranes, while UNet performed best on the carbon support film and antibody platforms. However, the loss values do not capture model performance in the same way as human judgement (Figure 2C). For the antibody platform, the model that would be expected to be one of the worst based on the loss values, Pix2pix, actually generates segmentations that are well-suited for the downstream processing tasks. It also outputs fewer false positive segmentations for sections of membranes than many other models, including the lowest-loss model UNet. Moreover, since Pix2pix is a relatively large network, it might also be improved further by increasing the number of training epochs. We thus decided to use Pix2pix for the segmentation of antibody platforms, and increased the size of the antibody platform training dataset (from 58 to 170 positive samples) to train a much improved second iteration of the network for use in the following analyses (Figure 2—figure supplement 1).

When taking the training and processing speeds into account as well as the segmentation results, there is no overall best architecture. We, therefore, included multiple well-performing model architectures in the final library, in order to allow users to select from these models to find one that works well for their specific datasets. Although it is not necessary to screen different network architectures and users may simply opt to use the default (VGGNet), these results thus show that it can be useful to test different networks to identify one that is best. Moreover, these results also highlight the utility of preparing well-performing models by iteratively improving training datasets and re-training models in a streamlined interface. To aid in this process, the software displays the loss value of a network during training and allows for the application of models to datasets during training. Thus, users can inspect how a model’s output changes during training and decide whether to interrupt training and improve the training data or choose a different architecture.

Fine-tuning segmentation results with model interactions

Although the above results go some way towards distinguishing the three different structures, they also show demonstrate a common limitation encountered in automated tomogram segmentation: some parts of the image are assigned a high segmentation value by multiple of the networks, leading to false classifications and ambiguity in the results. For example, the InceptionNet and ResNet antibody platform models falsely label the edges of the carbon film.

To further improve the segmentation results, we decided to implement a system of proximity-based ‘model interactions’ of two types, colocalization and avoidance (Figure 3A), using which the output of one model can be adjusted based on the output of other models. In a colocalization interaction, the predictions of one model (the child) are suppressed wherever the prediction value of another model (the parent) is below some threshold. In an avoidance interaction, suppression occurs wherever the parent model’s prediction value is above a threshold.

Figure 3

Download asset Open asset

These interactions are implemented as follows: first, a binary mask is generated by thresholding the parent model’s predictions using a user-specified threshold value. Next, the mask is then dilated using a circular kernel with a radius $R$, a parameter that we call the interaction radius. Finally, the child model’s prediction values are multiplied with this mask.

Besides these specific interactions between two models, the software also enables pitching multiple models against one another in what we call ‘model competition’. Models can be set to ‘emit’ and/or ‘absorb’ competition from other models. Here, to emit competition means that a model’s prediction value is included in a list of competing models. To absorb competition means that a model’s prediction value will be compared to all values in that list, and that this model’s prediction value for any pixel will be set to zero if any of the competing models’ prediction value is higher. On a pixel-by-pixel basis, all models that absorb competition are thus suppressed whenever their prediction value for a pixel is lower than that of any of the emitting models.

With the help of these model interactions, it is possible to suppress common erroneous segmentation results. For example, an interaction like ‘absorbing membrane model avoids emitting carbon model with R = 10 nm’ is effective at suppressing the prediction of edges of the carbon film as being membranes (Figure 3B). Another straightforward example of the utility of membrane interactions is the segmentation of membrane-bound particles. By defining the following two interactions: ‘antibody platform model avoids membrane model with R = 10 nm’ followed by ‘antibody platform model colocalizes with membrane model with R = 30 nm’, the Fc-platforms formed by IgG3 at a distance of ~22 nm from the membrane are retained, while false positive labelling of features such as the membrane or carbon is suppressed (Figure 3C).

By conditionally combining and editing the prediction results of multiple neural networks, model interactions can thus be helpful in fine-tuning segmentations to be better suited for downstream applications. To illustrate this, we generated a comparison of segmentation results using (i) no interactions, (ii) model competition only, and (iii) model competition as well as model interactions (Figure 3D), which demonstrates the degree to which false positives can be reduced by the use of model interactions (although at times at the expense of increasing the rate of false negatives).

Automating particle picking for subtomogram averaging

Protein structure determination by cryoET requires the careful selection of many subtomograms (i.e. sub-volumes of a tomogram that all contain the same structure of interest), and aligning and averaging these to generate a 3D reprojection of the structure of interest with a significantly increased signal to noise ratio. The process of selecting these sub-volumes is called ‘particle picking,’ and can be done either manually or in an automated fashion. Much time can be saved by automating particle picking based on segmentations. In many cases, though, segmentation results are not readily usable for particle picking, as they can often introduce numerous false positives. This is particularly the case with complex, feature-rich datasets such as those obtained within cells, where the structures of interest can visually appear highly similar to other structures that are also found in the data, or when the structures of interest are located close to other features and are, therefore, hard to isolate. An example of this latter case is the challenge of picking membrane-bound particles.

Recently, we have used cryoET and subtomogram averaging to determine the structures of membrane-bound IgG3 platforms and of IgG3 interacting with the human complement system component 1 (C1) on the surface of lipid vesicles (Abendstein et al., 2023). The reconstructions of the antibody platforms alone and of the antibody-C1 complex were prepared using 1193 and 2561 manually selected subtomograms, extracted from 55 and 101 tomograms, respectively. Manual picking of structures of interest, although very precise, is very time-consuming and in this particular case took approximately 20 hr of work.

To demonstrate the utility of our software for particle picking, we re-analyzed these same datasets, this time using Ais to automate the picking of the two structures: antibody platforms and antibody-C1 complexes. For the antibody platforms, we used the same models and model interactions as described above, while we trained an additional neural network to identify C1 complexes for the segmentation of the antibody-C1 complexes. To prepare a training dataset for this latter model, we opened all 101 tomograms in Ais and browsed the data to select and annotate slices where one or multiple antibody-C1 complexes were clearly visible (Figure 4A). The training dataset thus consisted of samples taken from multiple different tomograms; the annotation and data selection in this case took around 1 hr of work.

Figure 4 with 3 supplements see all

Download asset Open asset

The antibody platform and antibody-C1 complex models were then applied to the respective datasets, in combination with the membrane and carbon models and the model interactions described above (Figure 4B): the membrane avoiding carbon, and the antibody platforms colocalizing with the resulting membranes. We then launched a relatively large batch segmentation process (3 models × 156 tomograms) and left it to complete overnight.

Once complete, we used Ais to inspect the segmented volumes and original datasets in 3D, and adjusted the threshold value so that, in as far as possible, only the particles of interest remained visible and the number of false positive particles was minimized (Figure 4C). After selecting these values, we then launched a batch particle-picking process to determine lists of particle coordinates based on the segmented volumes.

Particle picking in Ais comprises a number of processing steps (Figure 4, Figure 4—figure supplement 1). First, the segmented (.mrc) volumes are thresholded at a user-specified level. Second, a distance transform of the resulting binary volume is computed, in which every nonzero pixel in the binary volume is assigned a new value, equal to the distance of that pixel to the nearest zero-valued pixel in the mask. Third, a watershed transform is applied to the resulting volume, so that the sets of pixels closest to any local maximum in the distance transformed volume are assigned to one group. Fourth, groups that are smaller than a user-specified minimum volume are discarded. Fifth, the remaining groups are assigned a weight value, equal to the sum of the prediction value (i.e. the corresponding pixel value in the input.mrc volume) of the pixels in the group. For every group found within close proximity to another group (using a user-specified value for the minimum particle spacing), the group with the lower weight value is discarded. Finally, the centroid coordinate of the grouped pixels is considered the final particle coordinate, and the list of all coordinates is saved in a tab-separated text file with values rounded to the nearest integer.

As an alternative output format, segmentations can also be converted to and saved as triangulated meshes (.obj file format), which can then be used for, e.g., membrane-guided particle picking (Pyle et al., 2022). After picking particles, the resulting coordinates are immediately available for inspection in the Ais 3D renderer (Figure 4, Figure 4—figure supplement 2).

After picking, we used EMAN2 (Galaz-Montoya et al., 2015, spt_boxer.py) to extract volumes using the particle coordinates as an input (Figure 4D), which resulted in 2499 vol for the antibody platform reconstruction and 602 for the antibody-C1 complex (n.b. these numbers can be highly dependent on the threshold value). These volumes, or subtomograms, were used as the input for subtomogram averaging using EMAN2 Galaz-Montoya et al., 2015 and Dynamo (Castaño-Díez et al., 2012; see Materials and methods), without further curation – i.e., we did not manually discard any of the extracted volumes.

After applying the same approach to subtomogram averaging as used previously (Abendstein et al., 2023), the resulting averages were indeed highly similar to the original reconstructions (Figure 4E and F, Figure 4—figure supplement 3). These results demonstrate that Ais can be successfully used to automate particle picking, and thus to significantly reduce the amount of time spent on what is often a laborious processing step.

Many-feature segmentations of complex in situ datasets

Aside from particle picking, segmentation is also often used to visualize and study the complex internal structure of a sample, as for example encountered when applying tomography to whole cells. Here too, the accuracy of a segmentation can be a critical factor in the success of downstream analyses such as performing measurements on the basis of 3D feature maps generated via segmentation.

A challenging aspect of the segmentation of cellular samples is that these datasets typically contain many features that are biologically distinct, but visually and computationally difficult to distinguish. For example, one challenge that is often encountered is that of distinguishing between various linearly shaped components: lipid membranes, actin filaments, microtubules, and intermediate filaments, which all appear as linear features with a relatively high density. To show the utility of Ais for the accurate segmentation of complex cellular tomograms, we next demonstrate a number of examples of such feature-rich segmentations.

The first example is a segmentation of seven distinct features observed in the base of Chlamydomonas reinhardtii cilia (Figure 5A), using the data by van den Hoek et al., 2022 that was deposited in the Electron Microscopy Public Image Archive (EMPIAR, Iudin et al., 2023) with accession number 11078. The features are: membranes, ribosomes, microtubule doublets, axial microtubules, non-microtubular filaments, interflagellar transport trains (IFTs), and glycocalyx. This dataset was particularly intricate (the supplementary information to the original publication lists more than 20 features that can be identified across the dataset) and some rare features, such as the IFTs, required careful annotation across all tomograms before we could compile a sufficiently large training dataset. The final segmentation correctly annotates most of the selected characteristics present in the sample: the ribosome exclusion zone that surrounds the ciliary base (van den Hoek et al., 2022) is clearly recognizable, and the structures of the glycocalyx, membranes, and microtubule doublets within the cilia are well defined. Some fractions of the meshwork of stellate fiber and Y-link proteins are also detected within the cilium.

Figure 5 with 2 supplements see all

Download asset Open asset

In the second example, we show a segmentation of six features found in and around a mitochondrion in a mouse neuron (Figure 5B), using the data by Wu et al., 2023 as available in the Electron Microscopy Data Bank (EMDB) (Lawson et al., 2016) with accession number 29207. In the original publication, the authors developed a segmentation method to detect and perform measurements on the granules found within mitochondria. Using Ais, we were able to prepare models to segment these granules, as well as microtubules, actin filaments, and ribosomes, and to distinguish between the highly similar vesicular membranes on the one hand, and the membranes of the mitochondrial cristae on the other.

Lastly, we used Ais to generate a 3D visualization of ten distinct cellular features observed in coronavirus-infected mammalian cells (Figure 5C, Figure 5—figure supplement 1), using data by Wolff et al., 2020: single membranes, double-membrane vesicles (DMVs), actin filaments, intermediate filaments, microtubules, mitochondrial granules, ribosomes, coronaviral nucleocapsid proteins, coronaviral pores in the DMVs, and the nucleic acids within the DMV replication organelles. The software was able to accurately distinguish between single membranes and double membranes, as well as to discriminate between the various filaments of the cytoskeleton. Moreover, we could identify the molecular pores within the DMV, and pick sets of particles that might be suitable for use in subtomogram averaging (see Figure 5, Figure 5—figure supplement 2). We could also segment the nucleocapsid proteins, thus distinguishing viral particles from other, similarly sized, single membrane vesicles, as well as detect the nucleic acids found within the DMVs. Although manual annotation took some hours in this case, the processing of the full volume took approximately 3 min per feature once the models were trained, and models can of course be applied to other volumes without requiring additional training.

To conclude, our aim with the development of Ais was to simplify and improve the accuracy of automated tomogram segmentation in order to make this processing step more accessible to many cryoET users. Here, we have attempted to create an intuitive and organized user interface that streamlines the whole workflow from annotation, to model preparation, to volume processing and particle picking and inspecting the results. Additionally, the model repository at aiscryoet.org is designed to aid users in achieving results even faster, by removing the need to generate custom models for common segmentation tasks. To help users become familiar with the software, documentation and tutorials are available at ais-cryoet.readthedocs.org and video tutorials can be accessed via youtube.com/@scNodes. By demonstrating the use of Ais in automating segmentation and particle picking for subtomogram averaging, and making the software available as an open-source project, we thus hope to help accelerate research and dissemination of data involving cryoET.

Source code is available via GitHub, (copy archived at Last, 2024), software available on the python package index as 'Ais-cryoET'. Trained neural networks have been uploaded to aiscryoet.org.

1. 1. Wu et al.

   (2023) Massive

   ID EMDB-29207. CryoET tomogram of mitochondria in BACHD mouse model neuron.

   https://www.ebi.ac.uk/emdb/EMD-29207
2. 1. Hoek V

   (2022) Massive

   ID EMPIAR-11078. In situ cryo-electron tomography of the C. reinhardtii ciliary transition zone.

   https://www.ebi.ac.uk/empiar/EMPIAR-11078/

1. Preprint

   1. Abadi M

   (2015) Large-Scale Machine Learning on Heterogeneous Distributed Systems

   arXiv.

   https://arxiv.org/abs/1603.04467

   • Google Scholar
2. 1. Abendstein L
   2. Dijkstra DJ
   3. Tjokrodirijo RTN
   4. van Veelen PA
   5. Trouw LA
   6. Hensbergen PJ
   7. Sharp TH

   (2023) Complement is activated by elevated IgG3 hexameric platforms and deposits C4b onto distinct antibody domains

   Nature Communications 14:4027.

   https://doi.org/10.1038/s41467-023-39788-5

   • PubMed
   • Google Scholar
3. 1. Bankhead P
   2. Loughrey MB
   3. Fernández JA
   4. Dombrowski Y
   5. McArt DG
   6. Dunne PD
   7. McQuaid S
   8. Gray RT
   9. Murray LJ
   10. Coleman HG
   11. James JA
   12. Salto-Tellez M
   13. Hamilton PW

   (2017) QuPath: Open source software for digital pathology image analysis

   Scientific Reports 7:16878.

   https://doi.org/10.1038/s41598-017-17204-5

   • PubMed
   • Google Scholar
4. 1. Belevich I
   2. Joensuu M
   3. Kumar D
   4. Vihinen H
   5. Jokitalo E

   (2016) Microscopy image browser: a platform for segmentation and analysis of multidimensional datasets

   PLOS Biology 14:e1002340.

   https://doi.org/10.1371/journal.pbio.1002340

   • PubMed
   • Google Scholar
5. 1. Burnley T
   2. Palmer CM
   3. Winn M

   (2017) Recent developments in the CCP-EM software suite

   Acta Crystallographica. Section D, Structural Biology 73:469–477.

   https://doi.org/10.1107/S2059798317007859

   • PubMed
   • Google Scholar
6. 1. Castaño-Díez D
   2. Kudryashev M
   3. Arheit M
   4. Stahlberg H

   (2012) Dynamo: a flexible, user-friendly development tool for subtomogram averaging of cryo-EM data in high-performance computing environments

   Journal of Structural Biology 178:139–151.

   https://doi.org/10.1016/j.jsb.2011.12.017

   • PubMed
   • Google Scholar
7. Software

   1. Cornut O

   (2023) Imgui, version 70b6ba4

   Github.

   https://github.com/ocornut/imgui
8. 1. Galaz-Montoya JG
   2. Flanagan J
   3. Schmid MF
   4. Ludtke SJ

   (2015) Single particle tomography in EMAN2

   Journal of Structural Biology 190:279–290.

   https://doi.org/10.1016/j.jsb.2015.04.016

   • PubMed
   • Google Scholar
9. 1. Goddard TD
   2. Huang CC
   3. Meng EC
   4. Pettersen EF
   5. Couch GS
   6. Morris JH
   7. Ferrin TE

   (2018) UCSF ChimeraX: Meeting modern challenges in visualization and analysis

   Protein Science 27:14–25.

   https://doi.org/10.1002/pro.3235

   • PubMed
   • Google Scholar
10. 1. Harris CR
    2. Millman KJ
    3. van der Walt SJ
    4. Gommers R
    5. Virtanen P
    6. Cournapeau D
    7. Wieser E
    8. Taylor J
    9. Berg S
    10. Smith NJ
    11. Kern R
    12. Picus M
    13. Hoyer S
    14. van Kerkwijk MH
    15. Brett M
    16. Haldane A
    17. Del Río JF
    18. Wiebe M
    19. Peterson P
    20. Gérard-Marchant P
    21. Sheppard K
    22. Reddy T
    23. Weckesser W
    24. Abbasi H
    25. Gohlke C
    26. Oliphant TE

    (2020) Array programming with NumPy

    Nature 585:357–362.

    https://doi.org/10.1038/s41586-020-2649-2

    • PubMed
    • Google Scholar
11. Conference

    1. Isola P
    2. Zhu JY
    3. Zhou T
    4. Efros AA

    (2017) Image-to-image translation with conditional adversarial networks

    2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR. pp. 5967–5976.

    https://doi.org/10.1109/CVPR.2017.632

    • Google Scholar
12. 1. Iudin A
    2. Korir PK
    3. Somasundharam S
    4. Weyand S
    5. Cattavitello C
    6. Fonseca N
    7. Salih O
    8. Kleywegt GJ
    9. Patwardhan A

    (2023) EMPIAR: the electron microscopy public image archive

    Nucleic Acids Research 51:D1503–D1511.

    https://doi.org/10.1093/nar/gkac1062

    • PubMed
    • Google Scholar
13. 1. Last MGF
    2. Voortman LM
    3. Sharp TH

    (2023) scNodes: a correlation and processing toolkit for super-resolution fluorescence and electron microscopy

    Nature Methods 20:1445–1446.

    https://doi.org/10.1038/s41592-023-01991-z

    • PubMed
    • Google Scholar
14. Software

    1. Last MGF

    (2024) Ais, version swh:1:rev:5494df88a67178b858f57b892603473c5b92115e

    Software Heritage.

    https://archive.softwareheritage.org/swh:1:dir:31a70077abc4603329f59153f10ed824af7e7727;origin=https://github.com/bionanopatterning/Ais;visit=swh:1:snp:07483a149be32b6fbef18df04edbe7698d9bc88e;anchor=swh:1:rev:5494df88a67178b858f57b892603473c5b92115e
15. 1. Lawson CL
    2. Patwardhan A
    3. Baker ML
    4. Hryc C
    5. Garcia ES
    6. Hudson BP
    7. Lagerstedt I
    8. Ludtke SJ
    9. Pintilie G
    10. Sala R
    11. Westbrook JD
    12. Berman HM
    13. Kleywegt GJ
    14. Chiu W

    (2016) EMDataBank unified data resource for 3DEM

    Nucleic Acids Research 44:D396–D403.

    https://doi.org/10.1093/nar/gkv1126

    • PubMed
    • Google Scholar
16. Software

    1. Linder-Norén E

    (2019) Keras-GAN, version 3ff3be4

    Github.

    https://github.com/eriklindernoren/Keras-GAN
17. 1. Luengo I
    2. Darrow MC
    3. Spink MC
    4. Sun Y
    5. Dai W
    6. He CY
    7. Chiu W
    8. Pridmore T
    9. Ashton AW
    10. Duke EMH
    11. Basham M
    12. French AP

    (2017) SuRVoS: super-region volume segmentation workbench

    Journal of Structural Biology 198:43–53.

    https://doi.org/10.1016/j.jsb.2017.02.007

    • PubMed
    • Google Scholar
18. 1. Mastronarde DN
    2. Held SR

    (2017) Automated tilt series alignment and tomographic reconstruction in IMOD

    Journal of Structural Biology 197:102–113.

    https://doi.org/10.1016/j.jsb.2016.07.011

    • PubMed
    • Google Scholar
19. 1. Pyle E
    2. Hutchings J
    3. Zanetti G

    (2022) Strategies for picking membrane-associated particles within subtomogram averaging workflows

    Faraday Discussions 240:101–113.

    https://doi.org/10.1039/d2fd00022a

    • PubMed
    • Google Scholar
20. Preprint

    1. Ronneberger O
    2. Fisher P
    3. Brox T

    (2015) U-Net: Convolutional Networks for Biomedical Image Segmentation

    arXiv.

    https://arxiv.org/abs/1505.04597

    • Google Scholar
21. Preprint

    1. Simonyan K
    2. Zisserman A

    (2015) Very Deep Convolutional Networks for Large-Scale Image Regognition

    arXiv.

    https://arxiv.org/abs/1409.1556

    • Google Scholar
22. Conference

    1. Szegedy C
    2. Sermanet P
    3. Reed S
    4. Anguelov D
    5. Erhan D
    6. Vanhoucke V
    7. Rabinovich A

    (2014) Going deeper with convolutions

    2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR.

    https://doi.org/10.1109/CVPR.2015.7298594

    • Google Scholar
23. Conference

    1. Szegedy C
    2. Ioffe S
    3. Vanhoucke V
    4. Alemi A

    (2016) Inception-v4, inception-ResNet and the impact of residual connections on learning

    Proceedings of the AAAI Conference on Artificial Intelligence.

    https://doi.org/10.1609/aaai.v31i1.11231

    • Google Scholar
24. 1. van den Hoek H
    2. Klena N
    3. Jordan MA
    4. Alvarez Viar G
    5. Righetto RD
    6. Schaffer M
    7. Erdmann PS
    8. Wan W
    9. Geimer S
    10. Plitzko JM
    11. Baumeister W
    12. Pigino G
    13. Hamel V
    14. Guichard P
    15. Engel BD

    (2022) In situ architecture of the ciliary base reveals the stepwise assembly of intraflagellar transport trains

    Science 377:543–548.

    https://doi.org/10.1126/science.abm6704

    • PubMed
    • Google Scholar
25. 1. van der Walt S
    2. Schönberger JL
    3. Nunez-Iglesias J
    4. Boulogne F
    5. Warner JD
    6. Yager N
    7. Gouillart E
    8. Yu T
    9. scikit-image contributors

    (2014) scikit-image: image processing in Python

    PeerJ 2:e453.

    https://doi.org/10.7717/peerj.453

    • PubMed
    • Google Scholar
26. 1. Virtanen P
    2. Gommers R
    3. Oliphant TE
    4. Haberland M
    5. Reddy T
    6. Cournapeau D
    7. Burovski E
    8. Peterson P
    9. Weckesser W
    10. Bright J
    11. van der Walt SJ
    12. Brett M
    13. Wilson J
    14. Millman KJ
    15. Mayorov N
    16. Nelson ARJ
    17. Jones E
    18. Kern R
    19. Larson E
    20. Carey CJ
    21. Polat İ
    22. Feng Y
    23. Moore EW
    24. VanderPlas J
    25. Laxalde D
    26. Perktold J
    27. Cimrman R
    28. Henriksen I
    29. Quintero EA
    30. Harris CR
    31. Archibald AM
    32. Ribeiro AH
    33. Pedregosa F
    34. van Mulbregt P
    35. SciPy 1.0 Contributors

    (2020) SciPy 1.0: fundamental algorithms for scientific computing in Python

    Nature Methods 17:261–272.

    https://doi.org/10.1038/s41592-019-0686-2

    • PubMed
    • Google Scholar
27. 1. Wolff G
    2. Limpens RWAL
    3. Zevenhoven-Dobbe JC
    4. Laugks U
    5. Zheng S
    6. de Jong AWM
    7. Koning RI
    8. Agard DA
    9. Grünewald K
    10. Koster AJ
    11. Snijder EJ
    12. Bárcena M

    (2020) A molecular pore spans the double membrane of the coronavirus replication organelle

    Science 369:1395–1398.

    https://doi.org/10.1126/science.abd3629

    • PubMed
    • Google Scholar
28. 1. Wu G-H
    2. Smith-Geater C
    3. Galaz-Montoya JG
    4. Gu Y
    5. Gupte SR
    6. Aviner R
    7. Mitchell PG
    8. Hsu J
    9. Miramontes R
    10. Wang KQ
    11. Geller NR
    12. Hou C
    13. Danita C
    14. Joubert L-M
    15. Schmid MF
    16. Yeung S
    17. Frydman J
    18. Mobley W
    19. Wu C
    20. Thompson LM
    21. Chiu W

    (2023) CryoET reveals organelle phenotypes in huntington disease patient iPSC-derived and mouse primary neurons

    Nature Communications 14:692.

    https://doi.org/10.1038/s41467-023-36096-w

    • PubMed
    • Google Scholar

Author details

1. Mart GF Last

   Department of Cell and Chemical Biology, Leiden University Medical Center, Leiden, Netherlands

   Contribution

   Conceptualization, Data curation, Software, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing – original draft, Writing – review and editing

   For correspondence

   mgflast@gmail.com

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-3739-8863

2. Leoni Abendstein

   1. Department of Cell and Chemical Biology, Leiden University Medical Center, Leiden, Netherlands
   2. Institute of Science and Technology Austria (ISTA), Klosterneuburg, Austria

   Contribution

   Resources, Data curation, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing – original draft, Writing – review and editing

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0001-7634-5353

3. Lenard M Voortman

   Department of Cell and Chemical Biology, Leiden University Medical Center, Leiden, Netherlands

   Contribution

   Supervision, Validation, Writing – original draft, Writing – review and editing

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0001-9794-067X

4. Thomas H Sharp

   1. Department of Cell and Chemical Biology, Leiden University Medical Center, Leiden, Netherlands
   2. School of Biochemistry, University of Bristol, Bristol, United Kingdom

   Contribution

   Supervision, Funding acquisition, Validation, Writing – original draft, Project administration, Writing – review and editing

   For correspondence

   t.sharp@bristol.ac.uk

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-1990-2333

• 1,029

  views

• 42

  downloads

• 0

  citations

Views, downloads and citations are aggregated across all versions of this paper published by eLife.