A Bayesian View on Cryo-EM Structure Determination

MRC Laboratory of Molecular Biology, Hills Road, Cambridge CB2 0QH, UK.
Journal of Molecular Biology (Impact Factor: 4.33). 11/2011; 415(2):406-18. DOI: 10.1016/j.jmb.2011.11.010
Source: PubMed

ABSTRACT Three-dimensional (3D) structure determination by single-particle analysis of cryo-electron microscopy (cryo-EM) images requires many parameters to be determined from extremely noisy data. This makes the method prone to overfitting, that is, when structures describe noise rather than signal, in particular near their resolution limit where noise levels are highest. Cryo-EM structures are typically filtered using ad hoc procedures to prevent overfitting, but the tuning of arbitrary parameters may lead to subjectivity in the results. I describe a Bayesian interpretation of cryo-EM structure determination, where smoothness in the reconstructed density is imposed through a Gaussian prior in the Fourier domain. The statistical framework dictates how data and prior knowledge should be combined, so that the optimal 3D linear filter is obtained without the need for arbitrariness and objective resolution estimates may be obtained. Application to experimental data indicates that the statistical approach yields more reliable structures than existing methods and is capable of detecting smaller classes in data sets that contain multiple different structures.

  • Source
    • "Relion version 1.2 was used for final refinement, resulting in a final 21-Å resolution structure (judging by the Fourier shell correlation [FSC] cutoff of 0.5) that clearly shows a ridge of density at the beak to hold it in the retracted position (Fig. 2 A and Fig. S1 B). Further multimodel 3D refinement and CTF correction in Relion (Scheres, 2012), which allows the data to separate into two distinct "
    [Show abstract] [Hide abstract]
    ABSTRACT: Casein kinase 1δ/ε (CK1δ/ε) and their yeast homologue Hrr25 are essential for cell growth. Further, CK1δ is overexpressed in several malignancies, and CK1δ inhibitors have shown promise in several preclinical animal studies. However, the substrates of Hrr25 and CK1δ/ε that are necessary for cell growth and survival are unknown. We show that Hrr25 is essential for ribosome assembly, where it phosphorylates the assembly factor Ltv1, which causes its release from nascent 40S subunits and allows subunit maturation. Hrr25 inactivation or expression of a nonphosphorylatable Ltv1 variant blocked Ltv1 release in vitro and in vivo, and prevented entry into the translation-like quality control cycle. Conversely, phosphomimetic Ltv1 variants rescued viability after Hrr25 depletion. Finally, Ltv1 knockdown in human breast cancer cells impaired apoptosis induced by CK1δ/ε inhibitors, establishing that the antiproliferative activity of these inhibitors is due, at least in part, to disruption of ribosome assembly. These findings validate the ribosome assembly pathway as a novel target for the development of anticancer therapeutics. © 2015 Ghalei et al.
    The Journal of Cell Biology 03/2015; 208(6):745-59. DOI:10.1083/jcb.201409056 · 9.69 Impact Factor
  • Source
    • "Finally, 219,169 and 160,264 particles, for low salt and high salt intermediates, respectively, were kept and subjected to single particle analysis. To explore the structural heterogeneity , classification was applied both at two dimensional (2D) and three dimensional (3D) levels, using RELION, an empirical Bayesian approach (Scheres, 2012). Both samples were first classified into 200 classes at 2D level in 25 iterations with an angle step of 5 degree, and bad particles were excluded from further analysis. "
    [Show abstract] [Hide abstract]
    ABSTRACT: The in vivo assembly of ribosomal subunits is a highly complex process, with a tight coordination between protein assembly and rRNA maturation events, such as folding and processing of rRNA precursors, as well as modifications of selected bases. In the cell, a large number of factors are required to ensure the efficiency and fidelity of subunit production. Here we characterize the immature 30S subunits accumulated in a factor-null Escherichia coli strain (∆rsgA∆rbfA). The immature 30S subunits isolated with varying salt concentrations in the buffer system show interesting differences on both protein composition and structure. Specifically, intermediates derived under the two contrasting salt conditions (high and low) likely reflect two distinctive assembly stages, the relatively early and late stages of the 3′ domain assembly, respectively. Detailed structural analysis demonstrates a mechanistic coupling between the maturation of the 5′ end of the 17S rRNA and the assembly of the 30S head domain, and attributes a unique role of S5 in coordinating these two events. Furthermore, our structural results likely reveal the location of the unprocessed terminal sequences of the 17S rRNA, and suggest that the maturation events of the 17S rRNA could be employed as quality control mechanisms on subunit production and protein translation. Electronic supplementary material The online version of this article (doi:10.1007/s13238-014-0044-1) contains supplementary material, which is available to authorized users.
    Protein & Cell 03/2014; 5(5). DOI:10.1007/s13238-014-0044-1 · 2.85 Impact Factor
  • Source
    • "However, their final numerical alternation procedure does not maximise the marginal likelihood, as an EM-ML would, but it maximises an approximate complete data likelihood instead. The EM-MAP approach [7] in the Fourier domain uses a smoothness prior for the particle. In addition, MAP estimation methods, not based on EM, have been proposed such as one based on sparsity [8]; and another computing the marginal posterior with a Monte Carlo estimate and a Gaussian smoothness prior for the reconstruction [9]. "
    [Show abstract] [Hide abstract]
    ABSTRACT: In this work, we address the problem of reconstructing the 3D structure of an object from a set of transmission electron microscopy (TEM) images, taken at unknown, random direc-tions around the object. We use the expectation maximisation (EM) algorithm for finding the maximum a posteriori (MAP) estimates for the 3D structure from the marginal posterior, where the view orientations are integrated out. In compari-son to previous work related to this single particle reconstruc-tion application, we have made the following novel contribu-tions. (1) We use Monte Carlo integration to approximate the expected complete data log posterior to reduce the computa-tional complexity; (2) we use a uniform prior in the space of rotations instead of the space of rotation angles; (3) we use the positivity constraint for the reconstructed density that is both a physical constraint as well as it acts as a natural sparsity prior; (4) on the M-step we use a large scale, subspace trust-region method based on the interior-reflective Newton method for efficient computation of the reconstruction. We experimented the approach on cryo-electron microscopy (cryo-EM) protein images. The results are promising and show that the 3D struc-ture can be robustly recovered with the proposed method in spite of the very low signal-to-noise ratio (SNR).
    International Symposium on Biomedical Imaging: From Nano to Macro, San Francisco, CA; 01/2013
Show more


Available from