Clustering and variance maps for cryo-electron tomography using wedge-masked differences

Abstract

Cryo-electron tomography provides 3D imaging of frozen hydrated biological samples with nanometer resolution. Reconstructed volumes suffer from low signal-to-noise-ratio (SNR)1 and artifacts caused by systematically missing tomographic data. Both problems can be overcome by combining multiple subvolumes with varying orientations, assuming they contain identical structures. Clustering (unsupervised classification) is required to ensure or verify population homogeneity, but this process is complicated by the problems of poor SNR and missing data, the factors that led to consideration of multiple subvolumes in the first place. Here, we describe a new approach to clustering and variance mapping in the face of these difficulties. The combined subvolume is taken as an estimate of the true subvolume, and the effect of missing data is computed for individual subvolumes. Clustering and variance mapping then proceed based on differences between expected and observed subvolumes. We show that this new method is faster and more accurate than two current, widely used techniques.

Architecture of respiratory syncytial virus revealed by electron cryotomography

“…Visual inspection of the M and RNP tomograms were carried out to investigate if there was any order in the data to justify subvolume averaging. All subvolume processing was carried out using 3dmod and PEET from the IMOD package (32)(33)(34). Bsoft was used for generating masks for subvolume processing and estimation of the subvolume average resolutions (35).…”

In Situ Localization of N and C Termini of Subunits of the Flagellar Nexin-Dynein Regulatory Complex (N-DRC) Using SNAP Tag and Cryo-electron Tomography

“…Subtomograms, each containing a 96-nm axonemal repeat, were extracted, aligned, and averaged with missing-wedge compensation using PEET software (3). An unsupervised classification approach incorporated in PEET (38) was used to analyze the presence or absence of strep-Au label in the axonemal repeats of WT, DRC4-SNAP, DRC3-SNAP, and SNAP-DRC3. During the classification analysis, the examined three-dimensional volume was limited to the targeted region with a mask.…”

PyTom: A python-based toolbox for localization of macromolecules in cryo-electron tomograms and subtomogram analysis

“…This dataset has been used in various studies of subtomogram alignment and classification and thus serves as a quasi-standard in the field (Förster et al, 2008;Heumann et al, 2011;Scheres et al, 2009;Yu and Frangakis, 2011). The 786 subtomograms were aligned using the average of the unaligned subtomograms as the initial reference.…”

“…Furthermore, classification techniques can be used to sort subtomograms according to their intrinsic variations, which yields an increase in resolution for the different classes compared to the overall average. Different software tools have been proposed for subtomogram localization (Förster et al, 2010;Frangakis and Forster, 2004;Rath et al, 2003;Renken et al, 2009;Xu et al, 2011), alignment (Amat et al, 2010;Bartesaghi et al, 2008;Förster et al, 2005;Schmid and Booth, 2008;Walz et al, 1997;Winkler, 2007) and classification (Bartesaghi et al, 2008;Förster et al, 2008;Heumann et al, 2011;Scheres et al, 2009;Stolken et al, 2010;Winkler et al, 2009;Yu et al, 2010;Yu and Frangakis, 2011), but a unified platform that covers the whole workflow from subtomogram localization to averaging and classification is not available to date. Here, we present the open-source toolbox PyTom, which covers a workflow comprising localization of macromolecules based on template matching, correlation-based alignment of subtomograms, and their classification by a novel stochastic method.…”