CryoDRGN: reconstruction of heterogeneous cryo-EM structures using neural networks

Zhong, Ellen D.; Bepler, Tristan; Berger, Bonnie; Davis, Joseph H.

doi:10.1038/s41592-020-01049-4

Cited by 397 publications

(531 citation statements)

References 53 publications

Supporting

Mentioning

466

Contrasting

Order By: Relevance

“…Additionally, performing cryo-EM on the 1-NO state may enable visualization of the different conformations present in the sample. In particular, advances in EM data processing techniques, such as the algorithm cryoDRGN, could facilitate the identification of discrete states and continuous conformational changes in heterogeneous samples [ 73 ]. Given these kinds of current, rapid advances in SAXS and EM analyses, there can be no doubt that understanding sGC from a structural perspective is only just beginning.…”

Section: Discussionmentioning

confidence: 99%

Structural Perspectives on the Mechanism of Soluble Guanylate Cyclase Activation

Wittenborn

Marletta

2021

IJMS

View full text Add to dashboard Cite

The enzyme soluble guanylate cyclase (sGC) is the prototypical nitric oxide (NO) receptor in humans and other higher eukaryotes and is responsible for transducing the initial NO signal to the secondary messenger cyclic guanosine monophosphate (cGMP). Generation of cGMP in turn leads to diverse physiological effects in the cardiopulmonary, vascular, and neurological systems. Given these important downstream effects, sGC has been biochemically characterized in great detail in the four decades since its discovery. Structures of full-length sGC, however, have proven elusive until very recently. In 2019, advances in single particle cryo–electron microscopy (cryo-EM) enabled visualization of full-length sGC for the first time. This review will summarize insights revealed by the structures of sGC in the unactivated and activated states and discuss their implications in the mechanism of sGC activation.

show abstract

Section: Discussionmentioning

confidence: 99%

Structural Perspectives on the Mechanism of Soluble Guanylate Cyclase Activation

Wittenborn

Marletta

2021

IJMS

View full text Add to dashboard Cite

show abstract

“…This issue becomes more important when other specimens of increased complexity are considered. A method that addresses this issue by approximating the continuous 3D density function of a single particle is CryoDRGN (Zhong et al, 2021), a deep neural network-based algorithm. Recent machine-learning methods may improve the protein density of experimental cryo-EM maps, while the use of generative adversarial networks (GANs) trained on pairs of 3D atomic models and their noise-free cryo-EM maps is shown to generate a more realistic ground-truth 3D density map (Sanchez-Garcia et al, 2020a).…”

Section: Processing (Cryo-)em Images From Native Extracts With a Focus On Machine Learningmentioning

confidence: 99%

Detecting Protein Communities in Native Cell Extracts by Machine Learning: A Structural Biologist’s Perspective

Kyrilis

Belapure

Kastritis

2021

Front. Mol. Biosci.

View full text Add to dashboard Cite

Native cell extracts hold great promise for understanding the molecular structure of ordered biological systems at high resolution. This is because higher-order biomolecular interactions, dubbed as protein communities, may be retained in their (near-)native state, in contrast to extensively purifying or artificially overexpressing the proteins of interest. The distinct machine-learning approaches are applied to discover protein–protein interactions within cell extracts, reconstruct dedicated biological networks, and report on protein community members from various organisms. Their validation is also important, e.g., by the cross-linking mass spectrometry or cell biology methods. In addition, the cell extracts are amenable to structural analysis by cryo-electron microscopy (cryo-EM), but due to their inherent complexity, sorting structural signatures of protein communities derived by cryo-EM comprises a formidable task. The application of image-processing workflows inspired by machine-learning techniques would provide improvements in distinguishing structural signatures, correlating proteomic and network data to structural signatures and subsequently reconstructed cryo-EM maps, and, ultimately, characterizing unidentified protein communities at high resolution. In this review article, we summarize recent literature in detecting protein communities from native cell extracts and identify the remaining challenges and opportunities. We argue that the progress in, and the integration of, machine learning, cryo-EM, and complementary structural proteomics approaches would provide the basis for a multi-scale molecular description of protein communities within native cell extracts.

show abstract

“…Principal component analysis of the Z values assigned to all particles revealed that the two opposite states along the first principal component of variability correspond to a state with ALC1 tightly bound to SHL2 and the H4 tail (same state described in Figure 2) and a state in which the ATPase domain is only loosely bound to the nucleosomal DNA. We used the graph traversal algorithm implemented in cryoDRGN (Zhong et al, 2021) with intermediate steps chosen to traverse the distribution along the first two principal components, which run approximately parallel to the two UMAP axes in this case (Figure 3A). Because cryo-EM data do not contain temporal information per se, this approach cannot inform on kinetics and temporal order.…”

Section: Analysis Of Heterogeneity In the Cryo-em Data Reveals Additional Functional States Of Alc1mentioning

confidence: 99%

“…Given the evidence that the set of 43 698 still contained significant continuous heterogeneity, we then analyzed it with cryoDRGN (Zhong et al, 2021). All models were trained for 50 epochs on a single GPU.…”

Section: Analysis Of Heterogeneity In the Cryo-em Datasetmentioning

confidence: 99%

Structure and dynamics of the chromatin remodeler ALC1 bound to a PARylated nucleosome

Bacic

Gaullier

Sabantsev

et al. 2021

Preprint

View full text Add to dashboard Cite

The chromatin remodeler ALC1 is recruited to and activated by DNA damage-induced poly(ADP-ribose) (PAR) chains deposited by PARP1/PARP2/HPF1 upon detection of DNA lesions. ALC1 has emerged as a candidate drug target for cancer therapy as its loss confers synthetic lethality in homologous recombination-deficient cells. However, structure-based drug design and molecular analysis of ALC1 have been hindered by the requirement for PARylation and the highly heterogeneous nature of this post-translational modification. Here, we reconstituted an ALC1 and PARylated nucleosome complex modified in vitro using PARP2 and HPF1. This complex was amenable to cryo-EM structure determination without cross-linking, which enabled visualization of several intermediate states of ALC1 from the recognition of the PARylated nucleosome to the tight binding and activation of the remodeler. Functional biochemical assays with PARylated nucleosomes highlight the importance of nucleosomal epitopes for productive remodeling and reveal that ALC1 preferentially slides nucleosomes away from DNA breaks.

show abstract

CryoDRGN: reconstruction of heterogeneous cryo-EM structures using neural networks

Cited by 397 publications

References 53 publications

Structural Perspectives on the Mechanism of Soluble Guanylate Cyclase Activation

Structural Perspectives on the Mechanism of Soluble Guanylate Cyclase Activation

Detecting Protein Communities in Native Cell Extracts by Machine Learning: A Structural Biologist’s Perspective

Structure and dynamics of the chromatin remodeler ALC1 bound to a PARylated nucleosome

Contact Info

Product

Resources

About