Convolutional neural networks for automated annotation of cellular cryo-electron tomograms

Chen, Muyuan; Dai, Wei; Sun, Stella; Jonasch, Darius; He, Cynthia Y.; Schmid, Michael F.; Chiu, Wah; Ludtke, Steven J.

doi:10.1038/nmeth.4405

Cited by 343 publications

(339 citation statements)

References 29 publications

Supporting

Mentioning

339

Contrasting

Order By: Relevance

“…8. Applying multiple binary networks is typically the strategy used in Chen et al 23 . In such a scenario, the network does not need to be as deep as multi-class networks, and accordingly can run faster while requiring less computing power.…”

Section: Handling Several Molecular Species Simultaneously Is Better mentioning

confidence: 99%

“…We demonstrate that manipulating several macromolecular species at the same time, is actually the key approach to improve performance of CNNs in 3D cryo-ET. Moreover, complex and time-consuming post-classification steps are no longer required to produce reliable results contrary to the aforementioned approaches 23 . We also show that DeepFinder is very flexible, and can be efficiently combined with template matching to improve localization sensitivity on experimental cryo-ET tomograms.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Deep Learning Improves Macromolecule Identification in 3D Cellular Cryo-Electron Tomograms

Moebel

Martínez-Sánchez

Lamm

et al. 2020

Preprint

View full text Add to dashboard Cite

Cryo-electron tomography (cryo-ET) allows one to visualize and study the 3D spatial distribution of macromolecules, in their native states and at nanometer resolution in single cells. While this label-free cryogenic imaging technology produces data containing rich structural information, automatic localization and identification of macromolecules are prone to noise and reconstruction artifacts, and to the presence of many molecular species in small areas. Hence, we present a computational procedure that uses artificial neural networks to accurately localize several macromolecular species in cellular cryo-electron tomograms. The DeepFinder algorithm leverages deep learning and outperforms the commonly-used template matching method on synthetic datasets. Meanwhile, DeepFinder is very fast when compared to template matching, and is better capable of localizing and identifying small macromolecules than other competitive deep learning methods. On experimental cryo-ET data depicting ribosomes, the localization and structure resolution (determined through subtomogram averaging) results obtained with DeepFinder are consistent with those obtained by experts. The DeepFinder algorithm is able to imitate the analysis performed by experts, and is therefore a very promising algorithm to investigate efficiently the contents of cellular tomograms. Furthermore, we show that DeepFinder can be combined with a template matching procedure to localize the missing macromolecules not found by one or the other method. Application of this collaborative strategy allowed us to find additional 20.5% membrane-bound ribosomes that had been missed or discarded during manual template matching-assisted annotation.

show abstract

Section: Handling Several Molecular Species Simultaneously Is Better mentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Deep Learning Improves Macromolecule Identification in 3D Cellular Cryo-Electron Tomograms

Moebel

Martínez-Sánchez

Lamm

et al. 2020

Preprint

View full text Add to dashboard Cite

show abstract

“…Even newer machine-learning algorithms for automated segmentation require that a user initially identify features manually within tomograms (Chen et al). …”

Section: Discussionmentioning

confidence: 99%

Exploring applications of crowdsourcing to cryo-EM

Bruggemann

Lander

2018

Journal of Structural Biology

View full text Add to dashboard Cite

Extraction of particles from cryo-electron microscopy (cryo-EM) micrographs is a crucial step in processing single-particle datasets. Although algorithms have been developed for automatic particle picking, these algorithms generally rely on two-dimensional templates for particle identification, which may exhibit biases that can propagate artifacts through the reconstruction pipeline. Manual picking is viewed as a gold-standard solution for particle selection, but it is too time-consuming to perform on data sets of thousands of images. In recent years, crowdsourcing has proven effective at leveraging the open web to manually curate datasets. In particular, citizen science projects such as Galaxy Zoo have shown the power of appealing to users’ scientific interests to process enormous amounts of data. To this end, we explored the possible applications of crowdsourcing in cryo-EM particle picking, presenting a variety of novel experiments including the production of a fully annotated particle set from untrained citizen scientists. We show the possibilities and limitations of crowdsourcing particle selection tasks, and explore further options for crowdsourcing cryo-EM data processing.

show abstract

“…Cryo-ET was used to determine the 3-dimensional organization of actin filaments bundled by dynamin. Krishna Chinthalapudi using the PDB entry 5ONV) into tomograms was done in Chimera 66 , and was also guided by segmented tomograms obtained using the EMAN2 semi-automated segmentation package 67 . The helical parameters of dynamin were calculated based on the diameter, pitch and known distance between dynamin subunits along the helical path.…”

Section: Cryo-electron Tomographymentioning

confidence: 99%

The mechanisms of dynamin-actin interaction

Zhang

Gerassimov

Lee

et al. 2019

Preprint

View full text Add to dashboard Cite

Cell-cell fusion is an indispensable process in the conception, development and physiology of multicellular organisms. Here we demonstrate a direct and noncanonical role for dynamin, best known as a fission GTPase in endocytosis, in cell-cell fusion. Our genetic and cell biological analyses show that dynamin colocalizes within the F-actin-enriched podosome-like structures at the fusogenic synapse, which is required for generating invasive membrane protrusions and myoblast fusion in vivo, in an endocytosis-independent manner. Biochemical, negative stain EM and cryo-electron tomography (cryo-ET) analyses revealed that dynamin forms helices that directly bundles actin filaments by capturing multiple actin filaments at their outer rim via interactions with dynamin's proline-rich domain.GTP hydrolysis by dynamin triggers disassembly of the dynamin helix, exposes the sides of the actin filaments, promotes dynamic Arp2/3-mediated branched actin polymerization, and generates a mechanically stiff actin network. Thus, dynamin functions as a unique actin-bundling protein that enhances mechanical force generation by the F-actin network in a GTPase-dependent manner. Our findings have universal implications for understanding dynamin-actin interactions in various cellular processes beyond cell-cell fusion. 3Cell-cell fusion is essential for the conception, development, regeneration and physiology of multicellular organisms 1-3 . A common theme underlying cell-cell fusion events ranging from insects to mammals is the presence of actin-propelled invasive protrusions at the site of fusion, known as the fusogenic synapse 4-8 . Such invasive protrusions were first identified as part of an F-actin-enriched podosome-like structure (PLS) in Drosophila myoblast fusion 7 . The PLS is generated by an "attacking" fusion partner, the fusion-competent myoblast (FCM), which drills multiple invasive protrusions into the "receiving" fusion partner, the muscle founder cell, to promote cell membrane juxtaposition and fusion pore formation 7,9,10 . Although each protrusion has a diameter similar to that of a filopodium, these protrusions are mechanically stiffer and are capable of triggering mechanosensitive responses in the receiving cell 11 . Mechanoresponsive accumulations of myosin II and spectrin in the receiving cell, in turn, increase the cortical tension of the receiving cell, thus generating resisting forces against the invasive protrusions and bringing the two cell membranes to close proximity to promote cell fusion 11,12 .The invasive protrusions from the attacking cells are propelled by Arp2/3-mediated branched actin polymerization, which generates short actin filaments more suitable for mechanical work than long and linear filaments 9 . However, how the branched actin filaments are organized to generate mechanically stiff membrane protrusions during fusion is unknown.Dynamin (Dyn) is a large GTPase best known for its role in endocytosis. Dynamin catalyzes membrane fission by forming rings/helices around necks of a budding endocytic ve...

show abstract

Convolutional neural networks for automated annotation of cellular cryo-electron tomograms

Cited by 343 publications

References 29 publications

Deep Learning Improves Macromolecule Identification in 3D Cellular Cryo-Electron Tomograms

Deep Learning Improves Macromolecule Identification in 3D Cellular Cryo-Electron Tomograms

Exploring applications of crowdsourcing to cryo-EM

The mechanisms of dynamin-actin interaction

Contact Info

Product

Resources

About