Multiparticle cryo-EM of ribosomes.

As the resolution of cryo-EM reconstructions has improved to the subnanometer range, conformational and compositional heterogeneity have become increasing problems in cryo-EM, limiting the resolution of reconstructions. Since further purification is not feasible, the presence of several conformational states of ribosomal complexes in thermodynamic equilibrium requires methods for separating these states in silico. We describe a procedure for generating subnanometer resolution cryo-EM structures from large sets of projection images of ribosomal complexes. The incremental K-means-like method of unsupervised 3D sorting discussed here allows separation of classes in the dataset by exploiting intrinsic divisions in the data. The classification procedure is described in detail and its effectiveness is illustrated using current examples from our work. Through a good separation of conformational modes, higher resolution reconstructions can be calculated. This increases information gained from single states, while exploiting the coexistence of multiple states to gather comprehensive mechanistic insight into biological processes like ribosomal translocation.