Hybrid Technologies for Structural Cell Biology at Near-Atomic Resolution
In 2020, cryo-electron tomography (cryo-ET) delivered a remarkable first 3.5Å resolution map of a macromolecular complex inside cells. While outstanding, this proof-of-principle was demonstrated on the ribosome, one of the largest and most abundant complexes in cells. To obtain high-resolution maps that reveal fine compositional and conformational states for any cellular complex, large data must be acquired, and all instances of the complex must be reliably localized within cryo-tomograms. This project will address the current technology shortcomings by synergizing cutting-edge technologies supported by the team’s expertise in methods development for in-cell structural biology, machine learning (ML), and super-resolution light microscopy.
Specifically, this team will harness ML to drive high-quality and high-throughput data acquisition on state-of-the-art microscopes. Researchers will develop cryo-single molecule localization microscopy (SMLM) to guide cryo-ET acquisition, and to localize small and low-abundance complexes in reference to easily detectable ribosomes. Ribosomes will inform on the considerable 3D specimen deformation during cryo-ET acquisition and guide precise 3D registration of cryo-ET and SMLM data. The team will also create a ML-based computational pipeline to analyse cryo-tomograms in a holistic approach by exploiting the cellular context, prior structural and biochemical knowledge, and the correlative cryo-SMLM data as weak supervision for complex detection and classification problems. They will demonstrate the feasibility of this hybrid approach for complexes that are not accessible with current technology in the genome-reduced bacterium Mycoplasma pneumoniae, and will explicitly construct the technology framework to allow transferable implementation for macromolecular complexes in higher complexity model systems.