The One World Cryo-EM seminar series is a platform for discussion of algorithms, computational methods and mathematical problems in cryo-EM.
 

Online, Every Other Wednesday, at 8am PT / 11am ET/ 4pm BST / 5pm CET / 11pm China.

Registration has changed. For zoom links and announcements, please register to our mailing list. This is a read-only mailing list (users cannot post). Please use your institution/ company email – we approve registrations manually and this helps us know who you are . You may receive an email asking you to confirm your registration, it is important that you click the link, otherwise we will not see your request.

Organizers: Joakim Andén, Dorit Hanein, Roy R. Lederman and Steven J. Ludtke


All TalksPast talksUpcoming Talks


Next Talk

Frank DiMaio

Associate Professor, Institute for Protein Design, University of Washington

Date and time: Wednesday September 7th, 2022 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.

Machine learning structure prediction and cryoEM map interpretation

The machine-learning structure prediction tools AlphaFold and RoseTTAfold have dramatically simplified low-resolution cryoEM map interpretation.  Following this, we describe three tools our lab has developed in conjunction with these approaches.  First, we describe novel machine-learning methodology for detecting errors in models built against low-resolution cryoEM density.  Next, we describe tools for modelling ligands into low-resolution cryoEM density.  Finally, we describe our efforts at adding to RoseTTAFold the ability to predict the structure of protein/nucleic-acid complexes.

Last Talk

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.

References

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).



Other “One Worlds” (not affiliated): https://www.owprobability.org/other-worlds