Figure 5 from Zhong et al 2021. a,b, Latent space representation of particle images of the assembling LSU (EMPIAR-10076) as a normalized histogram or UMAP embeddings after training a cryoDRGN 1D or 10D latent variable model, respectively. c,d, Latent space representation of particles colored by major LSU assembly state assigned from the 3D classification in Davis et al. Impurities in the dataset were assigned and subsequently filtered based on a cutoff of z = −1 in the 1D case (dotted line) and cluster assignment from a five-component Gaussian mixture model (GMM) in the 10D case. The dashed line in d indicates a rough outline of cluster assignment, shown in Extended Data Fig. 5. e, Density maps of the four major assembly states of the LSU reconstructed by cryoDRGN after training on the filtered dataset. Dashed lines indicate outlines of the fully mature 50S ribosome, with the central protuberance (CP) noted. f,g, Latent space representation of the filtered dataset, colored by major and minor assembly states assigned from the 3D classification in Davis et al. Points denote cluster centers for the corresponding assembly state. Major assembly state labels correspond to the structures from e. Inset shows a magnified view of the state C cluster and a population of particles originally misclassified into state E. h,i, CryoDRGN reconstruction of additional density maps, showing the 70S ribosome, an impurity during purification and LSU minor states C4 and E5. The newly identified C4 state resembles major state C in maturation but contains rRNA helix 68, previously present only in mature assembly states E4 and E5. j, Hyperparameters and runtime of the initial pilot experiments for particle filtering (a–d) and the final cryoDRGN model (e–i) trained on the assembling LSU dataset. Additional density maps are shown in Extended Data Fig. 6 and Supplementary Video 3.
