Determining the structure of proteins and other biomolecules at the atomic level is central to understanding many aspects of biology. X-ray crystallography is the best-known technique for structural biology but, as the name suggests, it works only with samples that can be crystallized. Electron cryo-microscopy (cryo-EM) could, potentially, be used to determine the atomic structures of biomolecules that cannot be crystallized, but at present the resolution that can be achieved with this approach is sufficient only for imaging certain types of viruses.

In cryo-EM, a solution of the biomolecule of interest is frozen in a thin layer of ice, and this layer is imaged in an electron microscope. By combining images of many identical biomolecules in many different orientations, it is possible to work backwards and determine their 3D structure. However, in order to determine this structure at high resolution, it is necessary to make repeated measurements to reduce high levels of noise in the images.

Cryo-EM images are usually recorded on a photographic film or a CCD (charge-coupled device) camera. However, photographic film is unsuitable for high-throughput methods because it has to be handled manually, while the efficiency of CCD cameras is limited because the electrons have to be converted into visible light to be detected. Digital cameras that can detect electrons directly have become available recently, and these are more efficient than both film and CCD cameras. They are also much faster, which means that it is possible to record videos of the sample during the time (typically ∼1 s) it is being exposed to the electron beam. Processing these videos could then—in theory—compensate for any movements of the biomolecules that are induced by the electron beam. Along with radiation damage caused by the electrons, these beam-induced movements have been a major limitation on the resolution that can be achieved with cryo-EM.

Bai et al. demonstrate the potential of direct-electron detectors in cryo-EM by determining the structures of two ribosomes. Using a novel statistical algorithm to accurately follow the movements of the ribosomes during the time they are exposed to the electron beam, they are able to compensate for these movements, and this makes it possible to determine the structures of the ribosomes with near-atomic precision. Moreover, the resolution they achieve with just ∼30,000 ribosomes is better than that previously achieved with more than a million ribosomes, allowing small details inside the ribosome – such as ß-strands and bulky amino-acid side chains – to be resolved with cryo-EM for the first time. The work of Bai et al. could, therefore, allow researchers to use cryo-EM to determine the structure of many more biomolecules with atomic precision.

For many years, electron cryo-microscopy (cryo-EM) has held the promise that macromolecular structure determination should in principle be possible to near-atomic resolution from only several thousand projection images of isolated particles in ice (Henderson, 1995; Glaeser, 1999). However, cryo-EM single-particle reconstructions in which protein main-chains may be traced reliably or individual side-chains are well resolved have been limited to large icosahedral viruses, and have typically required averaging over millions of asymmetric units (Grigorieff and Harrison, 2011). These results are in accordance with early observations that the contrast in EM images is not as good as expected from theory (Henderson, 1995). The main limiting factor of the information content in cryo-EM data is radiation damage: the electron dose needs to be limited to prevent the molecules from disintegrating while they are being imaged. However, radiation damage alone is not enough to account for the loss in expected contrast. Two additional factors that have been attributed to information loss in cryo-EM images are poor detective quantum efficiency (DQE, a frequency-dependent measure for the signal-to-noise performance) of conventional recording devices, and beam-induced movement or charging of the specimen during imaging (Henderson, 1992).

Traditionally, EM images have been recorded on photographic film, which has a large field of view and a reasonably good DQE, but is not convenient for high-throughput methods due to laborious steps of film development and digitization. Therefore, over the past decade many electron microscopes have been equipped with charge-coupled device (CCD) cameras. These digital detectors allow instant inspection of the images and are much more convenient in automated data collection schemes, so that large quantities of data may be obtained with relatively little effort (Stagg et al., 2006). However, the conversion of electrons into visible light that is detected by the CCD comes at the expense of a poorer DQE, in particular at higher voltages where many microscopes operate best. Consequently, despite the advantages of CCDs in terms of ease-of-use and data quantity, most of the near-atomic resolution cryo-EM structures published to date have been recorded on photographic film (Grigorieff and Harrison, 2011).

The second factor that has been attributed to the missing contrast in cryo-EM images is related to the radiation damage itself. For every elastic scattering event, which contributes positively to the phase contrast image, also three to four inelastic scattering events occur (Henderson, 1995). The inelastic events deposit energy in the sample, which leads to radiation damage by breaking covalent bonds and the generation of free radicals. Besides destroying the very structures one aims to determine, these interactions also generate charges in the sample that could lead to image blurring by deflecting the incoming electrons. In addition, radiolysis products such as hydrogen gas may build up internal pressure, thus leading to mechanical stress in the sample (Leapman and Sun, 1995). At high electron doses this stress ultimately leads to bubbling of the sample, while at much lower doses it has been postulated to induce movements that cause blurring of the images (Glaeser, 2008).

Recent technological advances may mitigate the image-blurring effects of both poor detectors and beam-induced movements. Building on technological developments that were initiated almost a decade ago (Prydderch et al., 2003), three different companies now sell digital cameras that detect electrons directly, that is, without the need to first convert electrons into visible light. Initial characterization of these direct electron detectors indicated that their DQE at high resolution is superior to that of CCD and film (McMullan et al., 2009a), in particular when backscattering of electrons is reduced by back-thinning of the substrate (McMullan et al., 2009c). An additional advantage of these devices is that they record images at high rates: ranging from 16 to 400 images per second for the currently available products. This allows the recording of a video during typical exposures, compared to a single image on film or CCD. By processing these videos, Grigorieff and colleagues previously showed that beam-induced movements could be followed for large icosahedral viruses, although the gain in resolution compared to similar data recorded on film was not as large as expected, possibly due to experimental design (Brilot et al., 2012; Campbell et al., 2012).

Here, we assess the potential of a back-thinned FEI Falcon direct electron detector for cryo-EM structure determination by recording videos on two well-characterized test samples: the prokaryotic and eukaryotic ribosome. Ribosome samples have been characterized by cryo-EM for more than 20 years (Frank et al., 1991), and encompass characteristics that are favorable to cryo-EM structure determination: their large content of dense RNA and their large molecular weight (2.8 MDa for prokaryotic 70S and ∼4 MDa for eukaryotic 80S ribosomes) result in images of relatively high contrast compared to many other macromolecular complexes, and RNA is known to be less radiation sensitive than protein. On the other hand, the ribosome is an intrinsically flexible macromolecular machine, which often leads to the presence of more than one structural state in ribosome samples. This represents a challenge, as advanced classification techniques are required to separate the different conformations into structurally homogeneous subsets that are amenable to single-particle reconstruction (Scheres et al., 2007; Fischer et al., 2010). In addition, the absence of symmetry in the ribosome makes it necessary to image many more particles (and to determine many more orientations) than is the case for icosahedral viruses with 60-fold symmetry. The combination of structural variability and the lack of symmetry may account for the fact that near-atomic resolution cryo-EM maps of ribosome structures have remained elusive thus far.

In the experiments described below, we used a 70S T. thermophilus ribosome sample to develop image-processing procedures that account for beam-induced movement of individual particles. Our detector operates at a rate of 17 frames per second, and we recorded 16-frame videos during 1-s exposures. At a total dose of 16 electrons/Ų, each individual video frame therefore integrates less than 1 electron/Ų. We expected the high noise levels in these individual frames to pose severe limitations on the accuracy with which these frames may be aligned. Because the alignment accuracy of individual particles remains unknown in typical single-particle analysis, we collected an entire data set as tilted pairs and used tilt-pair analysis to experimentally assess alignment errors for video frames (Rosenthal and Henderson, 2003). Guided by the results obtained from this analysis, we then developed a statistical video processing procedure that does not depend on tilted images. Application of this procedure to a second, untilted data set of 80S S. cerevisiae ribosomes serves to illustrate the impact these detectors, combined with suitable video processing procedures, will have on the field of cryo-EM structure determination.

We recorded tilt pairs of videos on the 70S T. thermophilus ribosome sample and used 2D and 3D classification to select a structurally homogeneous subpopulation of 15,202 particle pairs for 3D reconstruction (see ‘Materials and methods'). In the first instance, we disregarded the time-dimension of the videos, and used 16-frame averages for each particle (Figure 1). By assessing the consistency between the orientations from independently refined tilt pair particles with the known microscope tilt geometry, we estimated an alignment precision for the tilt pairs of 2° (Rosenthal and Henderson, 2003) (Figure 2A). This is a significant improvement over the 4° precision reported for a similar sample recorded on CCD (Henderson et al., 2011). This improvement reflects the improved signal-to-noise ratios (SNRs) in the individual particles, which are a direct consequence of the improved DQE of the detector. The benefits of using particles with higher SNRs for structure determination are therefore twofold. On the one hand, averaging images with higher SNRs requires fewer particles to overcome the noise. On the other hand, higher SNRs result in smaller alignment errors, which then lead to decreased blurring in the reconstruction. Both effects will lead to higher-resolution maps from fewer particles, which explains the relatively high resolution of 6.5 Å that we obtained from only 15,202 particle pairs.

Figure 1 with 1 supplement see all

Download asset Open asset

Figure 2 with 1 supplement see all

Download asset Open asset

We then assessed how accurately particle movement may be followed during the electron exposure. To this purpose, we partitioned each original 16-frame average particle into multiple independent particles that were calculated as the average of a varying number of individual video frames. In this manner, we artificially created four additional data sets of increasing size, comprising averages of 2 × 8, 3 × 6, 4 × 4 or 8 × 2 video frames for each original particle. Tilt-pair analysis for these data showed that, as expected, the numbers of correctly aligned particles as well as the alignment precision decrease with averaging over fewer video frames (Figure 2B). In addition, we noted a trend for particles that were calculated from later video frames to align less accurately, most likely as a result of accumulated radiation damage. To assess the effects of these alignment errors on reconstruction quality, we also inspected the Fourier Shell Correlation (FSC) curves that were reported for the four refinements (Figure 2C). The best reconstructions, at 5.4 Å resolution, were obtained when using averages of six or four individual video frames, which apparently represent the optimum between accurately describing particle movement and not introducing too large alignment errors for these data. Analysis of the orientations for the four-frame averaged particles that were aligned correctly according to the tilt-pair analysis revealed that overall rotations and translations during the exposure are in the order of 1.7 ± 1.2° and 4.2 ± 2.3 Å, respectively. Complicated patterns of movement for particles on some of the illuminated areas (e.g., see white lines in Figure 1A) indicate that the observed movements were not merely the result of a drifting stage, but more likely to be attributed to, as yet poorly understood, beam-induced effects.

Although useful for tilt-pair analysis, treating the averages of multiple video frames as completely independent also left valuable information unused. The observations that beam-induced movements are relatively small, and that the 16-frame averages may be aligned most accurately, represent valuable prior knowledge that may be efficiently expressed using statistical refinement methods. In the ‘Materials and methods' section we describe the implementation of a probabilistic prior on the orientations of the video frames inside a Bayesian refinement approach (Scheres, 2012a). This prior down-weights orientations for the video frames in a continuous manner according to their distance from the orientation determined for the 16-frame averages. Application of this procedure to the 70S ribosome data yielded a reconstruction with a resolution of 5.1 Å (orange line in Figure 2C). Comparison of the FSCs between the reconstructed maps and a rigid-body fitted crystal structure (Selmer et al., 2006), decreasing estimated B-factors, and visual inspection of the maps all confirmed that reconstruction quality may be improved by following the movement of individual particles, and that the statistical approach outperforms the straightforward use of multi-frame averaged particles in this respect (Figure 2D–F, Figure 2–figure supplement 1).

Finally, the developed procedure was applied to a second, untilted data set that was collected on an 80S S.cerevisiae ribosome sample during a single, manual microscopy session. Classification identified two major conformations, comprising 35,813 and 22,638 particles, respectively. A difference of ∼2° in ratchet-like movement represents the largest difference between the two classes (also see Video 1). Using the statistical video processing procedure, refinement of class 1 yielded a map with an overall resolution of 4.5 Å, while class 2 yielded a 4.6 Å map (Figure 3A). However, both maps still exhibited signs of unresolved structural heterogeneity, in particular in the small subunit and most notably for class 1 (Figure 3B). Therefore, the overall resolution estimates are too optimistic for the disordered parts, whereas the most stable parts of the structures still contain significant information beyond the overall estimated resolutions. The density for the 60S subunit clearly shows separation of β-strands, the pitch of α-helices, density for many side chains, and the separation of RNA bases, indicating that useful information to ∼4 Å resolution is present in the best parts of the map (Figure 3C–G, Video 2). The level of detail in the density for the 40S subunit varies, but is generally lower than for the 60S subunit (Figure 3H,I, Video 2). Calculation of FSC curves between the cryo-EM maps and rigid-body fitted crystal structures for the 60S and 40S subunits (Ben-Shem et al., 2011) confirmed the estimated resolutions, as well as the variation in quality of the density across the maps (Figure 3J,K).

Video 1

Download asset

Figure 3

Download asset Open asset

Video 2

Download asset

The reconstructions presented in this paper are of significantly higher resolution than any prokaryotic or eukaryotic ribosome structures that were calculated from cryo-EM data on photographic film or CCD. Moreover, they were obtained from unprecedented small numbers of particles. For comparison, currently the highest resolution ribosome map in the EMDB (an 80S structure from the plant Triticum aestivum) was calculated from almost 1.4 million particles and was reported to be at 5.5 Å resolution (Armache et al., 2010). We obtained reconstructions with useful information up to ∼4 Å resolution from nearly two orders of magnitude fewer particles. The observation that we could calculate maps with useful information up to 90% of the Nyquist frequency surpassed our initial expectations and clearly illustrates the potential of direct electron detectors combined with adequate video processing for cryo-EM structure determination.

Incorporation of the new detectors into automated data collection schemes will result in much larger amounts of data than we have used in our study. For relatively high molecular-weight complexes that display sufficient contrast for accurate alignment and classification, this may then allow high-resolution reconstruction of many more structural states, or of states that are adopted by only a small fraction of the particles. Moreover, if the sample is sufficiently rigid to exist in only one or a few distinct states, then larger amounts of data are expected to lead to even higher resolutions than the ones reported here. For example, following the SNR considerations proposed by (Rosenthal and Henderson, 2003), our overall resolution of 4.5 Å would increase to ∼3.3 Å if we would apply our approach to 1 million particles of similar quality (provided no other resolution-degrading factors play a role). Therefore, we foresee that the combination of direct electron detectors, video processing approaches, and automated data collection schemes will significantly increase the number of specimens for which de novo building of atomic models into cryo-EM maps becomes feasible.

The advantages of direct electron detectors and video processing procedures are not limited to relatively large complexes alone. Current alignment procedures are often severely limited by low SNRs in images of complexes that are smaller than ∼300 to 500 kDa. Also in this respect, the ability to record videos may provide significant benefits. Because SNRs in individual particles drop quickly with increasing spatial frequency image alignments are typically mainly driven by the lower-resolution components in the images (Henderson et al., 2011; Scheres and Chen, 2012). These components have been observed to remain intact after much higher electron doses than the high-resolution information (Hayward and Glaeser, 1979; Baker and Rubinstein, 2010). Therefore, relatively long videos with a high dose could be used to obtain reliable low-resolution components, while a dose-dependent weighting scheme to account for radiation damage may be used to optimize the SNR for higher spatial frequencies at the same time (Campbell et al., 2012). Such procedures are expected to significantly lower the size limit of particles that may still be aligned correctly and are thus amenable to cryo-EM structure determination. Still, accurately following the movement of small particles during videos will be more challenging than for the high-contrast ribosome particles presented here. Therefore, in order to also obtain near-atomic resolutions for smaller particles, further investigations into the nature of beam-induced image blurring and how to stop it will still be necessary. Also these investigations are expected to benefit significantly from the possibility to analyze beam-induced movements during videos recorded by direct electron detectors.

Based on our observations and the potential for further improvements in the technique, we believe the future for cryo-EM structure determination to be bright. Currently available detectors, combined with video processing algorithms like the one presented here, will result in higher-resolution cryo-EM structures for many more samples than those reported until now. Meanwhile, cryo-EM image quality will keep improving as ongoing technological developments such as single-electron counting detectors (McMullan et al., 2009b), phase-plate imaging (Nagayama, 2011) and better sample preparation techniques mature. Therefore, we are optimistic that the field will continue to progress towards fulfilling the promise of providing near-atomic resolution reconstructions, and thereby more detailed biological insights, for a wide range of specimens.

1. 1. Bai XC
   2. Fernandez IS
   3. McMullan G
   4. Scheres SHW

   (2013) Ribosome structures to near-atomic resolution from thirty thousand cryo-EM particles: S. cereviseae 80S ribosome

   ID 2275. In the public domain at the Electron Microscopy Data Bank (EMDB) at PDBE: http://www.ebi.ac.uk/pdbe/emdb/.

   http://www.ebi.ac.uk/pdbe-srv/emsearch/atlas/2275_summary.html
2. 1. Bai XC
   2. Fernandez IS
   3. McMullan G
   4. Scheres SHW

   (2013) Ribosome structures to near-atomic resolution from thirty thousand cryo-EM particles: S. cereviseae 80S ribosome

   ID 2276. In the public domain at the Electron Microscopy Data Bank (EMDB) at PDBE: http://www.ebi.ac.uk/pdbe/emdb/.

   http://www.ebi.ac.uk/pdbe-srv/emsearch/atlas/2276_sample.html
3. 1. Bai XC
   2. Fernandez IS
   3. McMullan G
   4. Scheres SHW

   (2013) Ribosome structures to near-atomic resolution from thirty thousand cryo-EM particles: T. thermophilus 70S ribosome

   ID 2277. In the public domain at the Electron Microscopy Data Bank (EMDB) at PDBE: http://www.ebi.ac.uk/pdbe/emdb/.

   http://www.ebi.ac.uk/pdbe-srv/emsearch/atlas/2277_summary.html
4. 1. Bai XC
   2. Fernandez IS
   3. McMullan G
   4. Scheres SHW

   (2013)

   Ribosome structures to near-atomic resolution from thirty thousand cryo-EM particles: 80S micrograph movies

   Unprocessed micrograph movies for the 80S sample will be distributed as 3DEM test data by the EMDB. Meanwhile, these data are available from the authors upon request.

1. 1. Armache J-P
   2. Jarasch A
   3. Anger AM
   4. Villa E
   5. Becker T
   6. Bhushan S

   et al. (2010) Cryo-EM structure and rRNA model of a translating eukaryotic 80S ribosome at 5.5-A resolution

   Proc Natl Acad Sci USA 107:19748–19753.

   https://doi.org/10.1073/pnas.1009999107

   • Google Scholar
2. 1. Baker LA
   2. Rubinstein JL

   (2010) Radiation damage in electron cryomicroscopy

   Methods Enzymol 481:371–388.

   https://doi.org/10.1016/S0076-6879(10)81015-8

   • Google Scholar
3. 1. Ben-Shem A
   2. Garreau de Loubresse N
   3. Melnikov S
   4. Jenner L
   5. Yusupova G
   6. Yusupov M

   (2011)

   The structure of the eukaryotic ribosome at 3.0 Å resolution

   Science 334:1524–1529.

   • Google Scholar
4. 1. Brilot AF
   2. Chen JZ
   3. Cheng A
   4. Pan J
   5. Harrison SC
   6. Potter CS

   et al. (2012) Beam-induced motion of vitrified specimen on holey carbon film

   J Struct Biol 177:630–637.

   https://doi.org/10.1016/j.jsb.2012.02.003

   • Google Scholar
5. 1. Campbell MG
   2. Cheng A
   3. Brilot AF
   4. Moeller A
   5. Lyumkis D
   6. Veesler D

   et al. (2012) Movies of ice-embedded particles enhance resolution in electron cryo-microscopy

   Structure 20:1823–1828.

   https://doi.org/10.1016/j.str.2012.08.026

   • Google Scholar
6. 1. Fischer N
   2. Konevega AL
   3. Wintermeyer W
   4. Rodnina MV
   5. Stark H

   (2010) Ribosome dynamics and tRNA movement by time-resolved electron cryomicroscopy

   Nature 466:329–333.

   https://doi.org/10.1038/nature09206

   • Google Scholar
7. 1. Frank J
   2. Penczek P
   3. Grassucci R
   4. Srivastava S

   (1991) Three-dimensional reconstruction of the 70S Escherichia coli ribosome in ice: the distribution of ribosomal RNA

   J Cell Biol 115:597–605.

   https://doi.org/10.1083/jcb.115.3.597

   • Google Scholar
8. 1. Glaeser RM

   (1999) Review: electron crystallography: present excitement, a nod to the past, anticipating the future

   J Struct Biol 128:3–14.

   https://doi.org/10.1006/jsbi.1999.4172

   • Google Scholar
9. 1. Glaeser RM

   (2008) Retrospective: radiation damage and its associated “information limitations”

   J Struct Biol 163:271–276.

   https://doi.org/10.1016/j.jsb.2008.06.001

   • Google Scholar
10. 1. Grigorieff N
    2. Harrison SC

    (2011) Near-atomic resolution reconstructions of icosahedral viruses from electron cryo-microscopy

    Curr Opin Struct Biol 21:265–273.

    https://doi.org/10.1016/j.sbi.2011.01.008

    • Google Scholar
11. 1. Hayward SB
    2. Glaeser RM

    (1979) Radiation damage of purple membrane at low temperature

    Ultramicroscopy 04:201–210.

    https://doi.org/10.1016/S0304-3991(79)90211-0

    • Google Scholar
12. 1. Henderson R

    (1992) Image contrast in high-resolution electron microscopy of biological macromolecules: TMV in ice

    Ultramicroscopy 46:1–18.

    https://doi.org/10.1016/0304-3991(92)90003-3

    • Google Scholar
13. 1. Henderson R

    (1995) The potential and limitations of neutrons, electrons and X-rays for atomic resolution microscopy of unstained biological molecules

    Q Rev Biophys 28:171–193.

    https://doi.org/10.1017/S003358350000305X

    • Google Scholar
14. 1. Henderson R
    2. Chen S
    3. Chen JZ
    4. Grigorieff N
    5. Passmore LA
    6. Ciccarelli L

    et al. (2011) Tilt-pair analysis of images from a range of different specimens in single-particle electron cryomicroscopy

    J Mol Biol 413:1028–1046.

    https://doi.org/10.1016/j.jmb.2011.09.008

    • Google Scholar
15. 1. Leapman RD
    2. Sun S

    (1995) Cryo-electron energy loss spectroscopy: observations on vitrified hydrated specimens and radiation damage

    Ultramicroscopy 59:71–79.

    https://doi.org/10.1016/0304-3991(95)00019-W

    • Google Scholar
16. 1. McMullan G
    2. Chen S
    3. Henderson R
    4. Faruqi AR

    (2009a) Detective quantum efficiency of electron area detectors in electron microscopy

    Ultramicroscopy 109:1126–1143.

    https://doi.org/10.1016/j.ultramic.2009.04.002

    • Google Scholar
17. 1. McMullan G
    2. Clark AT
    3. Turchetta R
    4. Faruqi AR

    (2009b) Enhanced imaging in low dose electron microscopy using electron counting

    Ultramicroscopy 109:1411–1416.

    https://doi.org/10.1016/j.ultramic.2009.07.004

    • Google Scholar
18. 1. McMullan G
    2. Faruqi AR
    3. Henderson R
    4. Guerrini N
    5. Turchetta R
    6. Jacobs A

    et al. (2009c) Experimental observation of the improvement in MTF from backthinning a CMOS direct electron detector

    Ultramicroscopy 109:1144–1147.

    https://doi.org/10.1016/j.ultramic.2009.05.005

    • Google Scholar
19. 1. Mindell JA
    2. Grigorieff N

    (2003) Accurate determination of local defocus and specimen tilt in electron microscopy

    J Struct Biol 142:334–347.

    https://doi.org/10.1016/S1047-8477(03)00069-8

    • Google Scholar
20. 1. Nagayama K

    (2011) Another 60 years in electron microscopy: development of phase-plate electron microscopy and biological applications

    J Electron Microsc 60:S43–S62.

    https://doi.org/10.1093/jmicro/dfr037

    • Google Scholar
21. 1. Pettersen EF
    2. Goddard TD
    3. Huang CC
    4. Couch GS
    5. Greenblatt DM
    6. Meng EC

    et al. (2004) UCSF Chimera—a visualization system for exploratory research and analysis

    J Comput Chem 25:1605–1612.

    https://doi.org/10.1002/jcc.20084

    • Google Scholar
22. 1. Prydderch M
    2. Waltham N
    3. Turchetta R
    4. French M
    5. Holt R
    6. Marshall A

    et al. (2003) A 512×512 CMOS monolithic active pixel sensor with integrated ADCs for space science

    Nucl Instrum Methods Phys Res A 512:358–367.

    https://doi.org/10.1016/S0168-9002(03)01914-4

    • Google Scholar
23. 1. Rosenthal PB
    2. Henderson R

    (2003)

    Optimal determination of particle orientation, absolute hand, and contrast loss in single-particle electron cryomicroscopy

    J Mol Biol 333:721–745.

    • Google Scholar
24. 1. Scheres SHW

    (2012a) A Bayesian view on cryo-EM structure determination

    J Mol Biol 415:406–418.

    https://doi.org/10.1016/j.jmb.2011.11.010

    • Google Scholar
25. 1. Scheres SHW

    (2012b) RELION: implementation of a Bayesian approach to cryo-EM structure determination

    J Struct Biol 180:519–530.

    https://doi.org/10.1016/j.jsb.2012.09.006

    • Google Scholar
26. 1. Scheres SHW
    2. Chen S

    (2012) Prevention of overfitting in cryo-EM structure determination

    Nat Methods 9:853–854.

    https://doi.org/10.1038/nmeth.2115

    • Google Scholar
27. 1. Scheres SHW
    2. Gao H
    3. Valle M
    4. Herman GT
    5. Eggermont PPB
    6. Frank J

    et al. (2007) Disentangling conformational states of macromolecules in 3D-EM through likelihood optimization

    Nat Methods 4:27–29.

    https://doi.org/10.1038/nmeth992

    • Google Scholar
28. 1. Seidelt B
    2. Innis CA
    3. Wilson DN
    4. Gartmann M
    5. Armache J-P
    6. Villa E

    et al. (2009) Structural insight into nascent polypeptide chain-mediated translational stalling

    Science 326:1412–1415.

    https://doi.org/10.1126/science.1177662

    • Google Scholar
29. 1. Selmer M
    2. Dunham CM
    3. Murphy FV IV
    4. Weixlbaumer A
    5. Petry S
    6. Kelley AC

    et al. (2006) Structure of the 70S ribosome complexed with mRNA and tRNA

    Science 313:1935–1942.

    https://doi.org/10.1126/science.1131127

    • Google Scholar
30. 1. Stagg SM
    2. Lander GC
    3. Pulokas J
    4. Fellmann D
    5. Cheng A
    6. Quispe JD

    et al. (2006) Automated cryoEM data acquisition and analysis of 284742 particles of GroEL

    J Struct Biol 155:470–481.

    https://doi.org/10.1016/j.jsb.2006.04.005

    • Google Scholar
31. 1. Tang G
    2. Peng L
    3. Baldwin PR
    4. Mann DS
    5. Jiang W
    6. Rees I

    et al. (2007) EMAN2: an extensible image processing suite for electron microscopy

    J Struct Biol 157:38–46.

    https://doi.org/10.1016/j.jsb.2006.05.009

    • Google Scholar
32. 1. Weixlbaumer A
    2. Jin H
    3. Neubauer C
    4. Voorhees RM
    5. Petry S
    6. Kelley AC

    et al. (2008) Insights into translational termination from the structure of RF2 bound to the ribosome

    Science 322:953–956.

    https://doi.org/10.1126/science.1164840

    • Google Scholar

Author details

1. Xiao-chen Bai

   Medical Research Council Laboratory of Molecular Biology, Cambridge, United Kingdom

   Contribution

   XB, Acquisition of data, Analysis and interpretation of data

   Competing interests

   The authors declare that no competing interest exist

2. Israel S Fernandez

   Medical Research Council Laboratory of Molecular Biology, Cambridge, United Kingdom

   Contribution

   ISF, Analysis and interpretation of data, Contributed unpublished essential data or reagents

   Competing interests

   The authors declare that no competing interest exist

3. Greg McMullan

   Medical Research Council Laboratory of Molecular Biology, Cambridge, United Kingdom

   Contribution

   GM, Acquisition of data

   Competing interests

   The authors declare that no competing interest exist

4. Sjors HW Scheres

   Medical Research Council Laboratory of Molecular Biology, Cambridge, United Kingdom

   Contribution

   SHWS, Conception and design, Analysis and interpretation of data, Drafting or revising the article

   For correspondence

   scheres@mrc-lmb.cam.ac.uk

   Competing interests

   The authors declare that no competing interest exist

• 13,806

  views

• 1,402

  downloads

• 363

  citations

Views, downloads and citations are aggregated across all versions of this paper published by eLife.