top of page

Deep Machine Learning for Quantifying Protein Dynamics from High-Speed Atomic Force Microscopy Data

December 11, 2020

Ziatdinov et al Highlight Graphic_Deep M

A deep learning (DL) network trained on a single labeled video frame is used to predict position and orientations of proteins in the entire movie (~600 frames) in a matter of seconds.

Scientific Achievement

A deep learning (DL) based workflow reveals the dynamics of protein self-assembly on inorganic substrates at the particle-by-particle level.

Significance and Impact

This work establishes the power of DL for the analysis of complex self-organization processes from advanced characterization data and provides insight into the fundamental mechanisms that underpin the system's behavior.

Research Details

  • The dynamics of self-assembly of a de novo designed protein were visualized using video-rate atomic force microscopy.

  • A DL network trained on a single frame was used for rapid conversion of the entire AFM movie into proteins position and orientation.

  • Possible classes of protein behaviors were identified, along with transition probabilities.

Ziatdinov, M., S. Zhang, O. Dollar, J. Pfaendtner, C.J. Mundy, X. Li, H. Pyles, D. Baker, J.J. De Yoreo, and S,V. Kalinin. (2021). Quantifying the dynamics of protein self-organization using deep learning analysis of atomic force microscopy data. Nano Letters 21: 158-165. DOI: 10.1021/acs.nanolett.0c03447

Work performed at the University of Washington, Pacific Northwest National Laboratory, and Oak Ridge National Laboratory

Thrust 1: Emergence of Order: Research


bottom of page