2022/06/08 Peter Schwander

Schwander. Peter

Peter Schwander

PhD, Assoc. Professor

Date and time: June 8th, 2022
Wednesday, at 8am PT / 11am ET / 4pm BST / 5pm CEST / 11pm China

For the zoom and gather.town links, please join the One World Cryo-EM mailing list.

Routes to Extract Biological Function from single-particle cryoEM 

Recent progress in data analysis of single-particle cryogenic Electron Microscopy (cryoEM); such as Geometric Machine Learning algorithms, allows one to retrieve the conformational spectrum of heterogeneous molecular ensembles.  Until recently, most efforts have been attempted under equilibrium conditions [1,2], where the conformational spectrum is time-independent, and thus directly yields the free-energy landscape. However, most biological systems operate far from equilibrium to sustain the processes that constitute life.

Under nonequilibrium conditions, the functional pathways are time-dependent, and the evolution of the conformational spectrum can be described by a Fokker-Planck equation, however with an unknown operator. State-of-the-art developments in Machine Learning, the so-called Physics-Informed Neural Networks (PINN) [3], allows one to retrieve the underlying Fokker-Planck operator from sparse observations alone [4], providing a complete physics-based description of the nonequilibrium process. Moreover, time-resolved cryoEM experiments have recently become practical [5]. Together, this enables us to combine the advantages of time-resolved serial crystallography (‘nonequilibrium processes’) with the advantages of single-particle methods (‘avoids averaging over unlike particles’).

Based on these opportunities, we present a conceptional and algorithmic framework to extract functional pathways from nonequilibrium from a collection of time-resolved single-particle images. This constitutes an unexplored route in studying biological function and structural dynamics under nonequilibrium conditions.


1. Dashti, A. et al. Trajectories of the ribosome as a Brownian nanomachine. Proc. Natl. Acad.  Sci. U. S. A. 111, 17492–7 (2014).

2. Dashti, A. et al. Retrieving functional pathways of biomolecules from single-particle snapshots. Nat. Commun. 11, 4734 (2020).

3. Raissi, M., Perdikaris, P. & Karniadakis, G. E. Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations. J. Comput. Phys. 378, (2019).

4. Chen, X., Yang, L., Duan, J. & Karniadakis, G. E. Solving inverse stochastic problems from discrete particle observations using the Fokker-Planck equation and physics-informed neural networks. SIAM Journal on Scientific Computing 43, (2021).

5. Dandey, V. et al. Time-resolved cryo-EM using Spotiton, Nature Meth., 17, 897 (2020).

2022/03/23 Sjors Scheres

Scheres, Sjors

Sjors Scheres

MRC Laboratory of Molecular Biology

Date and time: March 23rd, 2022
Wednesday, at 8am PT / 11am ET / 3pm GMT / 4pm CET / 11pm China

For the zoom and gather.town links, please join the One World Cryo-EM mailing list.

High-throughput cryo-EM structure determination of amyloids

The formation of amyloid filaments is characteristic of various degenerative diseases. Recent breakthroughs in electron cryo-microscopy (cryo-EM) have led to atomic structure determination of multiple amyloid filaments, both of filaments assembled in vitro from recombinant proteins, and of filaments extracted from diseased tissue. These observations revealed that a single protein may adopt multiple different amyloid folds, and that in vitro assembly does not necessarily lead to the same filaments as those observed in disease. In order to develop relevant model systems for disease, and ultimately to better understand the molecular mechanisms of disease, it will be important to determine which factors determine the formation of distinct amyloid folds. High-throughput cryo-EM methods will facilitate the screening of large numbers of in vitro assembly conditions. To this end, I will describe a new filament picking algorithm based on the Topaz approach, and outline image processing strategies in Relion that enable atomic structure determination of amyloids within days.

2022/01/26 Jonas Adler

Adler, Jonas

Jonas Adler

Senior Research Scientist at DeepMind

Date and time: January 26th, 2022
Wednesday, at 8am PT / 11am ET / 4pm GMT / 5pm CET / 12am China (Jan 27th)

For the zoom and gather.town links, please join the One World Cryo-EM mailing list.

Inferring a Continuous Distribution of Atom Coordinates from Cryo-EM Images using VAEs

Cryo-electron microscopy (cryo-EM) has revolutionized experimental protein structure determination. Despite advances in high resolution reconstruction, a majority of cryo-EM experiments provide either a single state of the studied macromolecule, or a relatively small number of its conformations. This reduces the effectiveness of the technique for proteins with flexible regions, which are known to play a key role in protein function. Recent methods for capturing conformational heterogeneity in cryo-EM data model it in volume space, making recovery of continuous atomic structures challenging. Here we present a fully deep-learning-based approach using variational auto-encoders (VAEs) to recover a continuous distribution of atomic protein structures and poses directly from picked particle images and demonstrate its efficacy on realistic simulated data. We hope that methods built on this work will allow incorporation of stronger prior information about protein structure and enable better understanding of non-rigid protein structures.

2021/12/01 Muyuan Chen

Chen, Muyuan

Muyuan Chen

Instructor, Baylor College of Medicine

Date and time: December 1st, 2021
Wednesday, at 8am PT / 11am ET / 4pm GMT / 5pm CET / 12am China (Dec 2nd)

For the zoom and gather.town links, please join the One World Cryo-EM mailing list.

Deep learning based Gaussian mixture model for characterizing variability in CryoEM

Structural flexibility and/or dynamic interactions with other molecules is a critical aspect of protein function. CryoEM provides direct visualization of individual macromolecules sampling different conformational and compositional states, but characterization of continuous conformational changes or large numbers of discrete states without human supervision remains challenging. Here, we present a machine learning algorithm to determine a conformational landscape for proteins or complexes using a 3-D Gaussian mixture model mapped onto 2-D particle images in known orientations. Using a deep neural network architecture, the method can automatically resolve the structural heterogeneity within the protein complex and map particles onto a small latent space describing conformational and compositional changes. The algorithm is applied to multiple biological systems, to explore compositional and conformational changes at a range of scales. 

2021/11/10 Harshit Gupta

Gupta, Harshit

Harshit Gupta

Postdoc, Stanford Linear Accelerator Center (SLAC)


Date and time: November 10th, 2021
Wednesday, at 8am PT / 11am ET / 4pm GMT / 5pm CET / 12am China (November 11th)

For the zoom and gather.town links, please join the One World Cryo-EM mailing list.

CryoGAN: A New Reconstruction Paradigm for Single-Particle Cryo-EM via Deep Adversarial Learning

Cryo-electron microscopy (Cryo-EM) has revolutionized structural biology over the last decade by delivering 3D structures of biomolecules at near-atomic resolution. It produces many noisy projections from separate instances of the same but randomly oriented biomolecule. These noisy projections are then used to reconstruct the 3D structure of the biomolecule. Scientists have spent the better part of the last 30 years designing a solid computational pipeline to achieve this goal. The result is an intricate multi-steps procedure that permits the regular discovery of new structures, but that is yet still prone to overfitting and irreproducibility. The most notable difficulties with the current paradigms are the need for pose-estimation methods, the reliance on user expertise for appropriate parameter tuning, and the non-straightforward extension to the handling of biomolecules with multiple conformations.

CryoGAN is a new paradigm for single-particle cryo-EM reconstruction based on unsupervised deep adversarial learning.  CryoGAN sidesteps the pose-estimation problem by using a generative adversarial network (GAN) to learn the 3D structure whose simulated projections most closely match the acquired projections in a distributional sense. The architecture of CryoGAN resembles that of standard GAN, with the twist that the generator network is replaced by a model of the cryo-EM image acquisition process. CryoGAN is an unsupervised algorithm that only demands projection images. No initial volume estimate or prior training is needed. CryoGAN requires minimal user interaction and can provide reconstructions in a matter of hours on a high-end GPU. In addition, it is backed by mathematical guarantees on the recovery of the correct structure. Moreover, its extension, called MultiCryoGAN can reconstruct continuous conformations of dynamic biomolecules, thus helping solve the most important open problem in the field without pose or conformation estimation.