Joseph (Joey) Davis: ”Visualizing ribosome assembly through cryo-EM, cryo-ET, and deep learning”

Macromolecular machines such as the ribosome undergo massivestructural changes as they assemble and function. While we have long appreciated such structural changes exist, experimentally visualizing and analyzing large ensembles of these structures is challenging. Here, I briefly describe cryoDRGN, a software package we developed to analyze structural heterogeneity in protein complexes visualized by cryo-electron microscopy (cryo-EM). This approach, which uses a purpose-built neural network based on a variational autoencoder, maps individual particle images to a low-dimensional latent space, effectively sorting particles based on their structure and allowing users to generate ensembles of hundreds-to-thousandsof three-dimensional structures using the trained networks. I detail our application of cryoDRGN to understand bacterial ribosome biogenesis generally and, specifically, how bacterial assembly cofactors including methyltransferases and RNA helicases guide assembly of the large and small ribosomal subunits.In analyzing a series of related cryo-EM datasets, we surprisingly uncovered that the methyltransferase KsgA ‘proof-reads’ the assembling ribosomes by preferentially disassembling ribosomes that have been erroneously constructed. Further, this work allowed us to visualize key assembly events that couple r-protein binding and rRNA folding, andto layer rRNA helical folding maps onto classical r-protein association maps established by Nomura, Nierhaus, and colleagues. Finally, I describe our work applying analogous methods we built to analyze structural heterogeneity in situ via cryo-electron tomography.Speaker Institution: Massachusetts Institute of Technology