CryoPoseNet: End-to-End Simultaneous Learning of Single-particle Orientation and 3D Map Reconstruction from Cryo-electron Microscopy Data

Nashed, Youssef S. G.; Poitevin, Frédéric; Gupta, Harshit; Woollard, Geoffrey; Kagan, M.; Yoon, Chun Hong; Ratner, Daniel

doi:10.1109/iccvw54120.2021.00452

Cited by 15 publications

(23 citation statements)

References 43 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Instead, each particle image is given a set of probable of orientations and similarity scores which eventually are used as weights in 3D reconstruction [ 34 , 36 , 47 , 50 , 51 ]. During each iteration, the estimation scores are improved until meeting a convergence criterion [ 51 , 52 ]. However, as it is still difficult to search all possible 3D maps, results heavily depend on the first estimate of the initial 3D model, resulting in an artifact known as model-bias [ 34 ].…”

Section: Three-dimensional (3d) Map Reconstructionmentioning

confidence: 99%

“…In particular, CNN-based models excel in image classification and particle recognition steps, which are the most fundamental steps in the cryo-EM image processing workflow [ 86 ]. Moreover, some neural-network algorithms capable of reconstructing high-resolution 3D structures for heterogenous samples are also proposed [ 24 , 25 , 52 ]. More recently, deep learning-based approaches related to post-processing, the final step associated with the enhancement of the reconstructed 3D electron density map, have also been launched [ 29 ].…”

Section: Future Applicationsmentioning

confidence: 99%

See 1 more Smart Citation

Artificial Intelligence in Cryo-Electron Microscopy

Chung

Durie

Lee

2022

Life

View full text Add to dashboard Cite

Cryo-electron microscopy (cryo-EM) has become an unrivaled tool for determining the structure of macromolecular complexes. The biological function of macromolecular complexes is inextricably tied to the flexibility of these complexes. Single particle cryo-EM can reveal the conformational heterogeneity of a biochemically pure sample, leading to well-founded mechanistic hypotheses about the roles these complexes play in biology. However, the processing of increasingly large, complex datasets using traditional data processing strategies is exceedingly expensive in both user time and computational resources. Current innovations in data processing capitalize on artificial intelligence (AI) to improve the efficiency of data analysis and validation. Here, we review new tools that use AI to automate the data analysis steps of particle picking, 3D map reconstruction, and local resolution determination. We discuss how the application of AI moves the field forward, and what obstacles remain. We also introduce potential future applications of AI to use cryo-EM in understanding protein communities in cells.

show abstract

Section: Three-dimensional (3d) Map Reconstructionmentioning

confidence: 99%

Section: Future Applicationsmentioning

confidence: 99%

Artificial Intelligence in Cryo-Electron Microscopy

Chung

Durie

Lee

2022

Life

View full text Add to dashboard Cite

show abstract

“…CryoPoseNet 24 and E2GMM 23 use an autoencoder, and E2GMM 23 additionally implements an variational autoencoder with a variational family of Gaussian distributions with fixed isotropic variance σ 0 . The AE architecture is used in CryoPoseNet 24 with a traditional L2 reconstruction loss and in E2GMM 23 with a tailored reconstruction loss that relies on the Fourier ring correlation (FRC) reconstruction metric.…”

Section: Ae and Vae In Cryodrgnmentioning

confidence: 99%

“…Non-Amortized Amortized (encoder) Non-Amortized Amortized (encoder) 19 (noise σ i ) (*) CryoPoseNet 24 (rotation R i ) FSTdiff 29 (rotation R i and conformation V ) CryoDRGN 15 (conformation z i ) CryoSPARC 10 (rotation R i ) (*) E2GMM 23 (conformation z i ) 3DFlex 22 (conformation z i ) CryoVAEGAN 32 (2D rotation R i and CT F i )…”

Section: /26mentioning

confidence: 99%

Deep Generative Modeling for Volume Reconstruction in Cryo-Electron Microscopy

Donnat¹,

Levy²,

Poitevin³

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

Recent breakthroughs in high resolution imaging of biomolecules in solution with cryo-electron microscopy (cryo-EM) have unlocked new doors for the reconstruction of molecular volumes, thereby promising further advances in biology, chemistry, and pharmacological research amongst others. Despite significant headway, the immense challenges in cryo-EM data analysis remain legion and intricately inter-disciplinary in nature, requiring insights from physicists, structural biologists, computer scientists, statisticians, and applied mathematicians. Meanwhile, recent next-generation volume reconstruction algorithms that combine generative modeling with end-to-end unsupervised deep learning techniques have shown promising results on simulated data, but still face considerable hurdles when applied to experimental cryo-EM images. In light of the proliferation of such methods and given the interdisciplinary nature of the task, we propose here a critical review of recent advances in the field of deep generative modeling for high resolution cryo-EM volume reconstruction. The present review aims to (i) compare and contrast these new methods, while (ii) presenting them from a perspective and using terminology familiar to scientists in each of the five aforementioned fields with no specific background in cryo-EM. The review begins with an introduction to the mathematical and computational challenges of deep generative models for cryo-EM volume reconstruction, along with an overview of the baseline methodology shared across this class of algorithms. Having established the common thread weaving through these different models, we then dive into the specifics and provide a practical comparison of these state-of-the-art algorithms, highlighting their relative strengths and weaknesses, along with the assumptions that they rely on. This allows us to identify bottlenecks in current methods and avenues for future research. This review might also raise the interest of computer vision practitioners, as it highlights current limits of deep generative models in a very challenging imaging regime where, for example, the signal-to-noise ratio is considerably lower than that found in traditional images.

show abstract

“…A more intelligent approach, called convolutional neural network (CNN), could provide a robust alternative solution. The CNN has great capabilities in feature extraction and nonlinear representation, allowing it outperform other algorithms in dealing with image-related tasks, such as image denoising, image reconstruction, − image classification, , and semantic segmentation. , Therefore, the CNN-based single image super-resolution (SISR) method , can achieve better performance regarding both the visual and quantitative aspect. Nonetheless, the implementation of SISR-CNN requires a large labeled data set to train the model, which is extremely unfriendly to the high-cost nanoscale imaging field.…”

Section: Introductionmentioning

confidence: 99%

Speeding up the Topography Imaging of Atomic Force Microscopy by Convolutional Neural Network

Zheng

Gao

et al. 2022

Anal. Chem.

View full text Add to dashboard Cite

Atomic force microscopy (AFM) provides unprecedented insight into surface topography research with ultrahigh spatial resolution at the subnanometer level. However, a slow scanning rate has to be employed to ensure the image quality, which will largely increase the accumulated sample drift, thereby, resulting in the low fidelity of the AFM image. In this paper, we propose a fast imaging method which performs a complete fast Raster scanning and a slow μ-path subsampling together with a deep learning algorithm to rapidly produce an AFM image with high quality and small drift. A supervised convolutional neural network (CNN) model is trained with the slow μ-path subsampled data and its counterpart acquired with fast Raster scan. The fast speed acquired AFM image is then inputted to the well-trained CNN model to output the high quality one. We validate the reliability of this method using a silicon grids sample and further apply it to the fast imaging of a vanadium dioxide thin film. The results demonstrate that this method can largely improve the imaging speed up to 10.3 times with state-of-the-art imaging quality, and reduce the sample drift by 8.9 times in the multiframe AFM imaging of the same area. Furthermore, we prove that this method is also applicable to other scanning imaging techniques such as scanning electrochemical microscopy.

show abstract

CryoPoseNet: End-to-End Simultaneous Learning of Single-particle Orientation and 3D Map Reconstruction from Cryo-electron Microscopy Data

Cited by 15 publications

References 43 publications

Artificial Intelligence in Cryo-Electron Microscopy

Artificial Intelligence in Cryo-Electron Microscopy

Deep Generative Modeling for Volume Reconstruction in Cryo-Electron Microscopy

Speeding up the Topography Imaging of Atomic Force Microscopy by Convolutional Neural Network

Contact Info

Product

Resources

About