TMaCS: A hybrid template matching and classification system for partially-automated particle selection

Selection of particle images from electron micrographs presents a bottleneck in determining the structures of macromolecular assemblies by single particle electron cryomicroscopy (cryo-EM). The problem is particularly important when an experimentalist wants to improve the resolution of a 3D map by increasing by tens or hundreds of thousands of images the size of the dataset used for calculating the map. Although several existing methods for automatic particle image selection work well for large protein complexes that produce high-contrast images, it is well known in the cryo-EM community that small complexes that give low-contrast images are often refractory to existing automated particle image selection schemes. Here we develop a method for partially-automated particle image selection when an initial 3D map of the protein under investigation is already available. Candidate particle images are selected from micrographs by template matching with template images derived from projections of the existing 3D map. The candidate particle images are then used to train a support vector machine, which classifies the candidates as particle images or non-particle images. In a final step in the analysis, the selected particle images are subjected to projection matching against the initial 3D map, with the correlation coefficient between the particle image and the best matching map projection used to assess the reliability of the particle image. We show that this approach is able to rapidly select particle images from micrographs of a rotary ATPase, a type of membrane protein complex involved in many aspects of biology.     

Auto-accumulation method using simulated annealing enables fully automatic particle pickup completely free from a matching template or learning data

Single-particle analysis is a 3-D structure determining method using electron microscopy (EM). In this method, a large number of projections is required to create 3-D reconstruction. In order to enable completely automatic pickup without a matching template or a training data set, we established a brand-new method in which the frames to pickup particles are randomly shifted and rotated over the electron micrograph and, using the total average image of the framed images as an index, each frame reaches a particle. In this process, shifts are selected to increase the contrast of the average. By iterated shifts and further selection of the shifts, the frames are induced to shift so as to surround particles. In this algorithm, hundreds of frames are initially distributed randomly over the electron micrograph in which multi-particle images are dispersed. Starting with these frames, one of them is selected and shifted randomly, and acceptance or non-acceptance of its new position is judged using the simulated annealing (SA) method in which the contrast score of the total average image is adopted as an index. After iteration of this process, the position of each frame converges so as to surround a particle and the framed images are picked up. This method is the first unsupervised fully automatic particle picking method which is applicable to EM of various kinds of proteins, especially to low-contrasted cryo-EM protein images.     

SLEUTH--a fast computer program for automatically detecting particles in electron microscope images

A method has been developed to locate biological complexes in a digitized electron micrograph by matching small windows to a set of reference images using a series of simple criteria. From the reference images, the program calculates parameters such as the radius of gyration, the density sum and variance. It compares them with corresponding values from a moving square window of densities extracted from the micrograph and records the coordinates of successfully matched candidate squares. Since the same particle is detected in a series of overlapping windows, candidates found to be within close proximity are grouped and the best-fitting one is selected from each cluster. The user is required only to select a small stack of boxed reference images and provide a few parameters, such as the particle radius and the minimum acceptable distance between particle centres. Micrograph labels and other areas that do not contain appropriate specimens are automatically ignored in order to minimize false positives. The program has been tested successfully on a variety of different biological structures, from both negatively stained and ice-embedded specimens.     

Accurate determination of local defocus and specimen tilt in electron microscopy

Accurate knowledge of defocus and tilt parameters is essential for the determination of three-dimensional protein structures at high resolution using electron microscopy. We present two computer programs, CTFFIND3 and CTFTILT, which determine defocus parameters from images of untilted specimens, as well as defocus and tilt parameters from images of tilted specimens, respectively. Both programs use a simple algorithm that fits the amplitude modulations visible in a power spectrum with a calculated contrast transfer function (CTF). The background present in the power spectrum is calculated using a low-pass filter. The background is then subtracted from the original power spectrum, allowing the fitting of only the oscillatory component of the CTF. CTFTILT determines specimen tilt parameters by measuring the defocus at a series of locations on the image while constraining them to a single plane. We tested the algorithm on images of two-dimensional crystals by comparing the results with those obtained using crystallographic methods. The images also contained contrast from carbon support film that added to the visibility of the CTF oscillations. The tests suggest that the fitting procedure is able to determine the image defocus with an error of about 10nm, whereas tilt axis and tilt angle are determined with an error of about 2 degrees and 1 degrees, respectively. Further tests were performed on images of single protein particles embedded in ice that were recorded from untilted or slightly tilted specimens. The visibility of the CTF oscillations from these images was reduced due to the lack of a carbon support film. Nevertheless, the test results suggest that the fitting procedure is able to determine image defocus and tilt angle with errors of about 100 nm and 6 degrees, respectively.     

FindEM--a fast, efficient program for automatic selection of particles from electron micrographs

The FindEM particle picking program was tested on the publicly available keyhole limpet hemocyanin (KLH) dataset, and the results were submitted for the "bakeoff" contest at the recent particle picking workshop (Zhu et al., 2003b). Two alternative ways of using the program are demonstrated and the results are compared. The first of these approximates exhaustive projection matching with a full set of expected views, which need to be known. This could correspond to the task of extending a known structure to higher resolution, for which many 1000's of additional images are required. The second procedure illustrates use of multivariate statistical analysis (MSA) to filter a preliminary set of candidate particles containing a high proportion of false particles. This set was generated using the FindEM program to search with one template that crudely represents the expected views. Classification of the resultant set of candidate particles then allows the desired classes to be selected while the rest can be ignored. This approach requires no prior information of the structure and is suitable for the initial investigation of an unknown structure--the class averages indicate the symmetry and oligomeric state of the particles. Potential improvements in speed and accuracy are discussed.     

Three-dimensional localization of immunogold markers using two tilted electron microscope recordings.

A method is presented to determine the three-dimensional positions of immuno-labeled gold markers from tilted electron micrograph recordings by using image processing techniques. The method consists of three basic modules: localization of the markers in the recordings, estimation of the motion parameters, and matching corresponding markers between the views. Localization consists of a segmentation step based on edge detection and region growing. It also allows for the separation of (visually) aggregated markers. Initial estimates for the motion parameters are obtained from a small number of user-indicated correspondences. A matching algorithm based on simulated annealing is used to find corresponding markers. With the resulting mapping, the motion parameters are updated and used in a new matching step, etc. Once the parameters are stable, the marker depths are retrieved. The developed method has been applied to semithin resin sections of A431 cells labeled for DNA and detected by silver-enhanced ultrasmall gold particles. It represents a reliable method to analyze the three-dimensional distribution of gold markers in electron microscope samples.     

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

A computational procedure is described for assigning the absolute hand of the structure of a protein or assembly determined by single-particle electron microscopy. The procedure requires a pair of micrographs of the same particle field recorded at two tilt angles of a single tilt-axis specimen holder together with the three-dimensional map whose hand is being determined. For orientations determined from particles on one micrograph using the map, the agreement (average phase residual) between particle images on the second micrograph and map projections is determined for all possible choices of tilt angle and axis. Whether the agreement is better at the known tilt angle and axis of the microscope or its inverse indicates whether the map is of correct or incorrect hand. An increased discrimination of correct from incorrect hand (free hand difference), as well as accurate identification of the known values for the tilt angle and axis, can be used as targets for rapidly optimizing the search or refinement procedures used to determine particle orientations. Optimized refinement reduces the tendency for the model to match noise in a single image, thus improving the accuracy of the orientation determination and therefore the quality of the resulting map. The hand determination and refinement optimization procedure is applied to image pairs of the dihydrolipoyl acetyltransferase (E2) catalytic core of the pyruvate dehydrogenase complex from Bacillus stearothermophilus taken by low-dose electron cryomicroscopy. Structure factor amplitudes of a three-dimensional map of the E2 catalytic core obtained by averaging untilted images of 3667 icosahedral particles are compared to a scattering reference using a Guinier plot. A noise-dependent structure factor weight is derived and used in conjunction with a temperature factor (B=-1000A(2)) to restore high-resolution contrast without amplifying noise and to visualize molecular features to 8.7A resolution, according to a new objective criterion for resolution assessment proposed here.     

Fast automatic particle picking from cryo-electron micrographs using a locally normalized cross-correlation function: a case study

Recent progress in single-particle reconstruction methods and cryo-EM techniques has led to the determination of macromolecular structures with unprecedented resolution. The number of particles that goes into the reconstruction is a key determinant in achieving high resolution. Interactive manual picking of particles from an electron micrograph is a very time-consuming, tedious, and inefficient process. We have implemented a fast automatic particle picking procedure in the SPIDER environment. The procedure makes use of template matching schemes and employs a recently developed locally normalized correlation algorithm based on Fourier techniques. As a test, we have used this procedure to pick 70S Escherichia coli ribosomes from a cryo-electron micrograph. Different search strategies including use of a circular mask and asymmetric masks for different orientations of the particle have been explored, and their relative efficiencies are discussed. The results indicate that the procedure can be optimally used to pick ribosomes in a fully automatic way within the limit of selecting less than 10% false positives while missing about 15% of true positives.     

Automatic CTF correction for single particles based upon multivariate statistical analysis of individual power spectra

Three-dimensional electron cryomicroscopy of randomly oriented single particles is a method that is suitable for the determination of three-dimensional structures of macromolecular complexes at molecular resolution. However, the electron-microscopical projection images are modulated by a contrast transfer function (CTF) that prevents the calculation of three-dimensional reconstructions of biological complexes at high resolution from uncorrected images. We describe here an automated method for the accurate determination and correction of the CTF parameters defocus, twofold astigmatism and amplitude-contrast proportion from single-particle images. At the same time, the method allows the frequency-dependent signal decrease (B factor) and the non-convoluted background signal to be estimated. The method involves the classification of the power spectra of single-particle images into groups with similar CTF parameters; this is done by multivariate statistical analysis (MSA) and hierarchically ascending classification (HAC). Averaging over several power spectra generates class averages with enhanced signal-to-noise ratios. The correct CTF parameters can be deduced from these class averages by applying an iterative correlation procedure with theoretical CTF functions; they are then used to correct the raw images. Furthermore, the method enables the tilt axis of the sample holder to be determined and allows the elimination of individual poor-quality images that show high drift or charging effects.     

EMAN: semiautomated software for high-resolution single-particle reconstructions

We present EMAN (Electron Micrograph ANalysis), a software package for performing semiautomated single-particle reconstructions from transmission electron micrographs. The goal of this project is to provide software capable of performing single-particle reconstructions beyond 10 A as such high-resolution data become available. A complete single-particle reconstruction algorithm is implemented. Options are available to generate an initial model for particles with no symmetry, a single axis of rotational symmetry, or icosahedral symmetry. Model refinement is an iterative process, which utilizes classification by model-based projection matching. CTF (contrast transfer function) parameters are determined using a new paradigm in which data from multiple micrographs are fit simultaneously. Amplitude and phase CTF correction is then performed automatically as part of the refinement loop. A graphical user interface is provided, so even those with little image processing experience will be able to begin performing reconstructions. Advanced users can directly use the lower level shell commands and even expand the package utilizing EMAN's extensive image-processing library. The package was written from scratch in C++ and is provided free of charge on our Web site. We present an overview of the package as well as several conformance tests with simulated data.     

Fast, robust, and accurate determination of transmission electron microscopy contrast transfer function

Transmission electron microscopy, as most imaging devices, introduces optical aberrations that in the case of thin specimens are usually modeled in Fourier space by the so-called contrast transfer function (CTF). Accurate determination of the CTF is crucial for its posterior correction. Furthermore, the CTF estimation must be fast and robust if high-throughput three-dimensional electron microscopy (3DEM) studies are to be carried out. In this paper we present a robust algorithm that fits a theoretical CTF model to the power spectrum density (PSD) measured on a specific micrograph or micrograph area. Our algorithm is capable of estimating the envelope of the CTF which is absolutely needed for the correction of the CTF amplitude changes.     

Automatic cryo-EM particle selection for membrane proteins in spherical liposomes

Random spherically constrained (RSC) single particle reconstruction is a method to obtain structures of membrane proteins embedded in lipid vesicles (liposomes). As in all single-particle cryo-EM methods, structure determination is greatly aided by reliable detection of protein “particles” in micrographs. After fitting and subtraction of the membrane density from a micrograph, normalized cross-correlation (NCC) and estimates of the particle signal amplitude are used to detect particles, using as references the projections of a 3D model. At each pixel position, the NCC is computed with only those references that are allowed by the geometric constraint of the particle’s embedding in the spherical vesicle membrane. We describe an efficient algorithm for computing this position-dependent correlation, and demonstrate its application to selection of membrane-protein particles, GluA2 glutamate receptors, which present very different views from different projection directions.     

Studies of the structure of the T4 bacteriophage tail sheath: I. The recovery of three-dimensional structural information from the extended sheath

The extended tail sheath of bacteriophage T4 has been used to study the transfer of information from an electron micrograph to the three-dimensional reconstruction obtained from it. Two methods have been developed to assess micrograph images of helical particles and their reconstructions. First, a filter has been designed which eliminates all structure in the image inconsistent with the symmetry and assumed radius of the helical particle. Individual micrographs can therefore be assessed with respect to their consistency with the assumed symmetry and radius, before reconstruction. Second, a map of the root-meansquare deviation of individual reconstructions from their average provides a quantitative measure of the consistency of the individual sets of tail data and allows the regions in the average reconstruction which are most sensitive to differences between the particles to be identified.The averaged reconstruction is used to examine the problems related to resolution and reproducibility of the structural information and to define the extent of the different components of the extended sheath.     

Image segmentation for automatic particle identification in electron micrographs based on hidden Markov random field models and expectation maximization

Three-dimensional reconstruction of large macromolecules like viruses at resolutions below 10 A requires a large set of projection images. Several automatic and semi-automatic particle detection algorithms have been developed along the years. Here we present a general technique designed to automatically identify the projection images of particles. The method is based on Markov random field modelling of the projected images and involves a pre-processing of electron micrographs followed by image segmentation and post-processing. The image is modelled as a coupling of two fields--a Markovian and a non-Markovian. The Markovian field represents the segmented image. The micrograph is the non-Markovian field. The image segmentation step involves an estimation of coupling parameters and the maximum á posteriori estimate of the realization of the Markovian field i.e, segmented image. Unlike most current methods, no bootstrapping with an initial selection of particles is required.     

Structural analysis of polymers of sickle cell hemoglobin. III. Fibers within fascicles

We have examined the structure of hemoglobin S fibers, which are associated into large bundles, or fascicles. Electron micrographs of embedded and cross-sectioned fascicles provide an end-on view of the component fibers. The cross-sectional images are rotationally blurred as a result of the twist of the fiber within the finite thickness of the section. We have applied restoration techniques to recover a deblurred image of the fiber. The first step in this procedure involved correlation averaging images of cross-sections of individual fibers in order to improve the signal-to-noise ratio. The rotationally blurred image was then geometrically transformed to polar co-ordinates. In this space, the rotational blur is transformed into a linear blur. The linearly blurred image is the convolution of the unblurred image and a point spread function that can be closely approximated by a square pulse. Deconvolution in Fourier space, followed by remapping to Cartesian co-ordinates, produced a deblurred image of the original micrograph. The deblurred images indicate that the fiber is comprised of 14 strands of hemoglobin S. This result provides confirmation of the fiber structure determined using helical reconstruction techniques and indicates that the association of fibers into ordered arrays does not alter their molecular structure.     

Single particle analysis of filamentous and highly elongated macromolecular assemblies

The application of single particle techniques to the three-dimensional analysis of electron microscope images of elongated or filamentous macromolecular assemblies is evaluated, taking as an example the muscle thin filament. Although the thin filament contains local helical symmetry, because of the inherent variable twist along it, the helical coherence does not extend for large enough distances to allow the symmetry to be used for full reconstruction of the tropomyosin/troponin repeat along the filament. The muscle thin filament therefore represents a general case of a filamentous object in that it is not possible to exploit symmetry in a full analysis. Due to the nature of the imaging process in the electron microscope, only projections of the thin filament around its long axis are available without tilting the grid. Crucially, projection images around a single axis do not provide enough information to assign Euler angles ab initio using current methods. Tests with a model thin filament structure indicated that an out-of-plane tilt of approximately 20 degrees was needed for ab initio angular assignment of sufficient accuracy to calculate a 3D structure to a resolution of approximately 25 A. If no out-of-plane views are available, an alternative approach is to use a prior 3D model as a reference for the initial angle assignment. Tests with the thin filament model indicated that reasonably accurate angular assignment can be made using a reference containing actin, but lacking the regulatory proteins tropomyosin and troponin. We also found that an adaptation of the exact filtered back projection method is required to allow the correct weighting of projection images in which the particle has a very large axial ratio. This adaptation resulted in significant improvements in the reconstruction.     

Regulation of CSF1 promoter by the SWI/SNF-like BAF complex

The mammalian BAF complex regulates gene expression by modifying chromatin structure. In this report, we identify 80 genes activated and 2 genes repressed by the BAF complex in SW-13 cells. We find that prior binding of NFI/CTF to the NFI/CTF binding site in CSF1 promoter is required for the recruitment of the BAF complex and the BAF-dependent activation of the promoter. Furthermore, the activation of the CSF1 promoter requires Z-DNA-forming sequences that are converted to Z-DNA structure upon activation by the BAF complex. The BAF complex facilitates Z-DNA formation in a nucleosomal template in vitro. We propose a model in which the BAF complex promotes Z-DNA formation which, in turn, stabilizes the open chromatin structure at the CSF1 promoter.     

Alignment error envelopes for single particle analysis

To determine the structure of a biological particle to high resolution by electron microscopy, image averaging is required to combine information from different views and to increase the signal-to-noise ratio. Starting from the number of noiseless views necessary to resolve features of a given size, four general factors are considered that increase the number of images actually needed: (1) the physics of electron scattering introduces shot noise, (2) thermal motion and particle inhomogeneity cause the scattered electrons to describe a mixture of structures, (3) the microscope system fails to usefully record all the information carried by the scattered electrons, and (4) image misalignment leads to information loss through incoherent averaging. The compound effect of factors 2-4 is approximated by the product of envelope functions. The problem of incoherent image averaging is developed in detail through derivation of five envelope functions that account for small errors in 11 "alignment" parameters describing particle location, orientation, defocus, magnification, and beam tilt. The analysis provides target error tolerances for single particle analysis to near-atomic (3.5 A) resolution, and this prospect is shown to depend critically on image quality, defocus determination, and microscope alignment.     

Effect of number of views on cross-sectional Compton imaging: A fundamental study with backprojection

IntroductionWe have been developing a medical imaging technique using a Compton camera, which is expected to reconstruct three-dimensional images. If the number of views is not sufficient, star-shaped artifacts (streak artifacts) could arise in cross-sectional images. Therefore, we estimated the point spread function (PSF) of cross-sectional Compton images and the effect of the number of views by Monte Carlo simulations and experimental studies.Materials and methodsA cross-sectional Compton image was reconstructed using a dataset comprising 719 view directions and PSF was analyzed using a radial distribution. The peak height, full width at half maximum (FWHM), background (BG), and residual sum of squares (RSS) were calculated from the obtained PSF. In addition, RSSs were plotted against the number of views to estimate the required number to suppress star-shaped artifacts.ResultsThere was no correlation found between the number of views and both FWHM (16 mm) and peak/BG ratio (∼1 × 10⁴). RSSs were reduced with the number of views and approached the minimum asymptotically. Correlation was observed between the required number of views and the number of Compton events used for image reconstruction.ConclusionWe determined the PSF of cross-sectional Compton images and the effect of the number of views on the images. The required number of views to suppress the star-shaped artifact is related to the square root of the number of Compton events used to reconstruct the image. From this study, we concluded that 21 or more views are required for clinical purposes.