Unsupervised Deep Representation Learning and Foundation Models for Cryo-Electron Tomography Analysis
Center for Structural Biology Seminar

Abstract: Cellular cryo-electron tomography (cryo-ET) offers the unprecedented ability to visualize the near-native 3D spatial organization of subcellular structures directly within single cells. However, the systematic analysis of these structures is fundamentally bottlenecked by extreme signal-to-noise limitations, the missing wedge effect, and the manual annotation burdens inherently associated with existing methodologies. To overcome these analytical barriers, our research program has explored approaches based on unsupervised and self-supervised deep learning methods and foundation models. Collectively, these approaches contribute toward translating massive, unannotated cryo-ET datasets into interpretable subcellular organization.

Contact

  Katsuhiko Murakami
  kum14@psu.edu
  +1 814-865-2758