A machine learning approach for the identification of protein secondary structure elements from electron cryo-microscopy density maps

The accuracy of the secondary structure element (SSE) identification from volumetric protein density maps is critical for de-novo backbone structure derivation in electron cryo-microscopy (cryoEM). It is still challenging to detect the SSE automatically and accurately from the density maps at medium resolutions (～5-10 ?). We present a machine learning approach, SSELearner, to automatically identify helices and β-sheets by using the knowledge from existing volumetric maps in the Electron Microscopy Data Bank. We tested our approach using 10 simulated density maps. The averaged specificity and sensitivity for the helix detection are 94.9% and 95.8%, respectively, and those for the β-sheet detection are 86.7% and 96.4%, respectively. We have developed a secondary structure annotator, SSID, to predict the helices and β-strands from the backbone Cα trace. With the help of SSID, we tested our SSELearner using 13 experimentally derived cryo-EM density maps. The machine learning approach shows the specificity and sensitivity of 91.8% and 74.5%, respectively, for the helix detection and 85.2% and 86.5% respectively for the β-sheet detection in cryoEM maps of Electron Microscopy Data Bank. The reduced detection accuracy reveals the challenges in SSE detection when the cryoEM maps are used instead of the simulated maps. Our results suggest that it is effective to use one cryoEM map for learning to detect the SSE in another cryoEM map of similar quality.     

Automatic determination of the handedness of single-particle maps of macromolecules solved by CryoEM

Single-Particle Analysis by Cryo-Electron Microscopy is a well-established technique to elucidate the three-dimensional (3D) structure of biological macromolecules. The orientation of the acquired projection images must be initially estimated without any reference to the final structure. In this step, algorithms may find a mirrored version of all the orientations resulting in a mirrored 3D map. It is as compatible with the acquired images as its unmirrored version from the image processing point of view, only that it is not biologically plausible.In this article, we introduce HaPi (Handedness Pipeline), the first method to automatically determine the hand of electron density maps of macromolecules solved by CryoEM. HaPi is built by training two 3D convolutional neural networks. The first determines α-helices in a map, and the second determines whether the α-helix is left-handed or right-handed. A consensus strategy defines the overall map hand. The pipeline is trained on simulated and experimental data. The handedness can be detected only for maps whose resolution is better than 5 Å. HaPi can identify the hand in 89% of new simulated maps correctly. Moreover, we evaluated all the maps deposited at the Electron Microscopy Data Bank and 11 structures uploaded with the incorrect hand were identified.     

Common conventions for interchange and archiving of three-dimensional electron microscopy information in structural biology

Three-dimensional electron microscopy (3DEM) has made significant contributions to structural biology. To accomplish this feat, many image-processing software packages were developed by various laboratories. The independent development of methods naturally implied the adoption of dissimilar conventions-penalizing users who want to take advantage of the wealth of algorithms from different packages. In addition, a public repository of 3DEM research results, the EM Data Bank, is now established. In an era where information exchange is important, standardizing conventions is a necessity. The 3DEM field requires a consistent set of conventions. We propose a set of common conventions named the "3DEM Image Conventions." They are designed as a standardized approach to image interpretation and presentation. In this regard, the conventions serve as a first step on which to build data-exchange solutions among existing software packages and as a vehicle for homogenous data representation in data archives, such as the EM Data Bank.     

Combining electron microscopy and comparative protein structure modeling

Recently, advances have been made in methods and applications that integrate electron microscopy density maps and comparative modeling to produce atomic structures of macromolecular assemblies. Electron microscopy can benefit from comparative modeling through the fitting of comparative models into electron microscopy density maps. Also, comparative modeling can benefit from electron microscopy through the use of intermediate-resolution density maps in fold recognition, template selection and sequence-structure alignment.     

A high-level 3D visualization API for Java and ImageJ

Background  

Current imaging methods such as Magnetic Resonance Imaging (MRI), Confocal microscopy, Electron Microscopy (EM) or Selective Plane Illumination Microscopy (SPIM) yield three-dimensional (3D) data sets in need of appropriate computational methods for their analysis. The reconstruction, segmentation and registration are best approached from the 3D representation of the data set.     

Computational approaches for automatic structural analysis of large biomolecular complexes

We present computational solutions to two problemsof macromolecular structure interpretation from reconstructedthree-dimensional electron microscopy (3D-EM) maps of largebio-molecular complexes at intermediate resolution (5A-15A). Thetwo problems addressed are: (a) 3D structural alignment (matching)between identified and segmented 3D maps of structure units(e.g. trimeric configuration of proteins), and (b) the secondarystructure identification of a segmented protein 3D map (i.e.locations of a-helices, b -sheets). For problem (a), we presentan efficient algorithm to correlate spatially (and structurally)two 3D maps of structure units. Besides providing a similarityscore between structure units, the algorithm yields an effectivetechnique for resolution refinement of repeated structure units,by 3D alignment and averaging. For problem (b), we present anefficient algorithm to compute eigenvalues and link eigenvectorsof a Gaussian convoluted structure tensor derived from theprotein 3D Map, thereby identifying and locating secondarystructural motifs of proteins. The efficiency and performanceof our approach is demonstrated on several experimentallyreconstructed 3D maps of virus capsid shells from single-particlecryo-EM, as well as computationally simulated protein structuredensity 3D maps generated from protein model entries in theProtein Data Bank.     

Imaging Transient Blood Vessel Fusion Events in Zebrafish by Correlative Volume Electron Microscopy

The study of biological processes has become increasingly reliant on obtaining high-resolution spatial and temporal data through imaging techniques. As researchers demand molecular resolution of cellular events in the context of whole organisms, correlation of non-invasive live-organism imaging with electron microscopy in complex three-dimensional samples becomes critical. The developing blood vessels of vertebrates form a highly complex network which cannot be imaged at high resolution using traditional methods. Here we show that the point of fusion between growing blood vessels of transgenic zebrafish, identified in live confocal microscopy, can subsequently be traced through the structure of the organism using Focused Ion Beam/Scanning Electron Microscopy (FIB/SEM) and Serial Block Face/Scanning Electron Microscopy (SBF/SEM). The resulting data give unprecedented microanatomical detail of the zebrafish and, for the first time, allow visualization of the ultrastructure of a time-limited biological event within the context of a whole organism.     

A novel three-dimensional variant of the watershed transform for segmentation of electron density maps

Electron density maps at moderate resolution are often difficult to interpret due to the lack of recognizable features. This is especially true for electron tomograms that suffer in addition to the resolution limitation from low signal-to-noise ratios. Reliable segmentation of such maps into smaller, manageable units can greatly facilitate interpretation. Here, we present a segmentation approach targeting three-dimensional electron density maps derived by electron microscopy. The approach consists of a novel three-dimensional variant of the immersion-based watershed algorithm. We tested the algorithm on calculated data and applied it to a wide variety of electron density maps ranging from reconstructions of single macromolecules to tomograms of subcellular structures. The results indicate that the algorithm is reliable, efficient, accurate, and applicable to a wide variety of biological problems.     

Crystal structure of human immunoglobulin fragment Fab New refined at 2.0 A resolution.

The three-dimensional structure of the human immunoglobulin fragment Fab New (IgG1, lambda) has been refined to a crystallographic R-factor of 16.9% to 2 A resolution. Rms deviations of the final model from ideal geometry are 0.014 A for bond distances and 3.03 degrees for bond angles. Refinement was based on a new X-ray data set including 28,301 reflections with F > 2.5 sigma(F) from 6.0 to 2.0 A resolution. The starting model for the refinement procedure reported here is from the Brookhaven Protein Data Bank entry 3FAB (rev. 1981). Differences between the initial and final models include modified polypeptide-chain folding in the third complementarity-determining region (CDR3) and the third framework region (FR3) of VH and in some exposed loops of CL and CH1. Amino acid sequence changes were determined at a number of positions by inspection of difference electron density maps. The incorporation of amino acid sequence changes results in an improved VH framework model for the "humanization" of monoclonal antibodies.     

Lipid-membrane protein interaction visualised by cryo-EM: A review

Membrane proteins reside at interfaces between aqueous and lipid media and solving their molecular structure relies most of the time on removing them from the membrane using detergent. Luckily, this solubilization process does not strip them from all the associated lipids and single-particle cryo-transmission electron microscopy (SP-TEM) has proved a very good tool to visualise both protein high-resolution structure and, often, many of its associated lipids. In this review, we observe membrane protein structures from the Protein DataBank and their associated maps in the Electron Microscopy DataBase and determine how the SP-TEM maps allow lipid visualization, the type of binding sites, the influence of symmetry, resolution and other factors. We illustrate lipid visualization around and inside the protein core, show that some lipid bilayers in the core can be shifted with respect to the membrane and how some proteins can actively bend the lipid bilayer that binds to them. We conclude that resolution improvement in SP-TEM will likely enable many more discoveries regarding the role of lipids bound to proteins.     

Studying intracellular transport using high-pressure freezing and Correlative Light Electron Microscopy

Correlative Light Electron Microscopy (CLEM) aims at combining the best of light and electron microscopy in one experiment. Light microscopy (LM) is especially suited for providing a general overview with data from lots of different cells and by using live cell imaging it can show the history or sequence of events between or inside cells. Electron microscopy (EM) on the other hand can provide a much higher resolution image of a particular event and provide additional spatial information, the so-called reference space. CLEM thus has certain strengths over the application of both LM and EM techniques separately. But combining both modalities however generally also means making compromises in one or both of the techniques. Most often the preservation of ultrastructure for the electron microscopy part is sacrificed. Ideally samples should be visualized in its most native state both in the light microscope as well as the electron microscope. For electron microscopy this currently means that the sample will have to be cryo-fixed instead of the standard chemical fixation. In this paper we will discuss the rationale for using cryofixation for CLEM experiments. In particular we will highlight a CLEM technique using high-pressure freezing in combination with live cell imaging. In addition we examine some of the EM analysis tools that may be useful in combination with CLEM techniques.     

A quantitative investigation of microvascular changes in the thyroid gland after infrared (IR) laser radiation.

We present an ultrastructural study of thyroid capillaries in which 50-day-old rats Wistar rats, were irradiated with an infrared (IR) laser, (total dose, 46.80 J/cm2), the tissue quantified 1 day after ending treatment and a quantitative capillary analysis carried out by light and electron microscopy. Light microscopy was used to calculate capillary volume density revealing a significant increase in the irradiated rats. The quantitative measurement of parameters by electron microscopy required a two stage analysis: Level I, Electron Microscopy (Magnification x5,000); and Level II, Electron Microscopy (Magnification x26,000). At Level I, the following parameters were measured in each capillary: capillary area, capillary diameter, luminal area, luminal diameter, endothelial area, nuclear area and mean endothelial thickness. At Level II, pinocytotic vesicle diameter and their numerical density in endothelial cells were evaluated. Electron microscopic analysis revealed an increased luminal area in the capillaries of the irradiated rats. They also presented a decrease in endothelial cell thickness and vesicular diameter and an increase in vesicle numerical density. This latter increase is indicative of presumptive changes in capillary permeability, but the possible functional significance of these morphological changes in the endothelial cells requires further investigation.     

Estimating loop length from CryoEM images at medium resolutions

Background

De novo protein modeling approaches utilize 3-dimensional (3D) images derived from electron cryomicroscopy (CryoEM) experiments. The skeleton connecting two secondary structures such as α-helices represent the loop in the 3D image. The accuracy of the skeleton and of the detected secondary structures are critical in De novo modeling. It is important to measure the length along the skeleton accurately since the length can be used as a constraint in modeling the protein.

Results

We have developed a novel computational geometric approach to derive a simplified curve in order to estimate the loop length along the skeleton. The method was tested using fifty simulated density images of helix-loop-helix segments of atomic structures and eighteen experimentally derived density data from Electron Microscopy Data Bank (EMDB). The test using simulated density maps shows that it is possible to estimate within 0.5Å of the expected length for 48 of the 50 cases. The experiments, involving eighteen experimentally derived CryoEM images, show that twelve cases have error within 2Å.

Conclusions

The tests using both simulated and experimentally derived images show that it is possible for our proposed method to estimate the loop length along the skeleton if the secondary structure elements, such as α-helices, can be detected accurately, and there is a continuous skeleton linking the α-helices.

     

Protein structure databases with new web services for structural biology and biomedical research

The Protein Data Bank Japan (PDBj) curates, edits and distributes protein structural data as a member of the worldwide Protein Data Bank (wwPDB) and currently processes approximately 25-30% of all deposited data in the world. Structural information is enhanced by the addition of biological and biochemical functional data as well as experimental details extracted from the literature and other databases. Several applications have been developed at PDBj for structural biology and biomedical studies: (i) a Java-based molecular graphics viewer, jV; (ii) display of electron density maps for the evaluation of structure quality; (iii) an extensive database of molecular surfaces for functional sites, eF-site, as well as a search service for similar molecular surfaces, eF-seek; (iv) identification of sequence and structural neighbors; (v) a graphical user interface to all known protein folds with links to the above applications, Protein Globe. Recent examples are shown that highlight the utility of these tools in recognizing remote homologies between pairs of protein structures and in assigning putative biochemical functions to newly determined targets from structural genomics projects.     

Conformational dynamics of supramolecular protein assemblies

Supramolecular protein assemblies including molecular motors, cytoskeletal filaments, chaperones, and ribosomes play a central role in a broad array of cellular functions ranging from cell division and motility to RNA and protein synthesis and folding. Single-particle reconstructions of such assemblies have been growing rapidly in recent years, providing increasingly high resolution structural information under native conditions. While the static structure of these assemblies provides essential insight into their mechanism of biological function, their dynamical motions provide additional important information that cannot be inferred from structure alone. Here we present an unsupervised computational framework for the analysis of high molecular weight protein assemblies and use it to analyze the conformational dynamics of structures deposited in the Electron Microscopy Data Bank. Protein assemblies are modeled using a recently introduced coarse-grained modeling framework based on the finite element method, which is used to compute equilibrium thermal fluctuations, elastic strain energy distributions associated with specific conformational transitions, and dynamical correlations in distant molecular domains. Results are presented in detail for the ribosome-bound termination factor RF2 from Escherichia coli, the nuclear pore complex from Dictyostelium discoideum, and the chaperonin GroEL from E. coli. Elastic strain energy distributions reveal hinge-regions associated with specific conformational change pathways, and correlations in collective molecular motions reveal dynamical coupling between distant molecular domains that suggest new, as well as confirm existing, allosteric mechanisms. Results are publically available for use in further investigation and interpretation of biological function including cooperative transitions, allosteric communication, and molecular mechanics, as well as in further classification and refinement of electron microscopy based structures.     

Deep learning for reconstructing protein structures from cryo-EM density maps: Recent advances and future directions

Cryo-Electron Microscopy (cryo-EM) has emerged as a key technology to determine the structure of proteins, particularly large protein complexes and assemblies in recent years. A key challenge in cryo-EM data analysis is to automatically reconstruct accurate protein structures from cryo-EM density maps. In this review, we briefly overview various deep learning methods for building protein structures from cryo-EM density maps, analyze their impact, and discuss the challenges of preparing high-quality data sets for training deep learning models. Looking into the future, more advanced deep learning models of effectively integrating cryo-EM data with other sources of complementary data such as protein sequences and AlphaFold-predicted structures need to be developed to further advance the field.     

Fingerprint-based structure retrieval using electron density

We present a computational approach that can quickly search a large protein structural database to identify structures that fit a given electron density, such as determined by cryo-electron microscopy. We use geometric invariants (fingerprints) constructed using 3D Zernike moments to describe the electron density, and reduce the problem of fitting of the structure to the electron density to simple fingerprint comparison. Using this approach, we are able to screen the entire Protein Data Bank and identify structures that fit two experimental electron densities determined by cryo-electron microscopy.     

Re-examining the spectra of macromolecules. Current practice of spectral quasi B-factor flattening

The analysis of structure factors in 3D cryo-EM Coulomb potential maps and their “enhancement” at the end of the reconstruction process is a well-established practice, normally referred to as sharpening. The aim is to increase contrast and, in this way, to help tracing the atomic model. The most common way to accomplish this enhancement is by means of the so-called B-factor correction, which applies a global filter to boost high frequencies with some dampening considerations related to noise amplification. The results are maps with a better visual aspect and a quasiflat spectrum at medium and high frequencies. This practice is so widespread that most map depositions in the Electron Microscopy Data Base (EMDB) only contain sharpened maps. Here, the use in cryoEM of global B-factor corrections is theoretically and experimentally analyzed. Results clearly illustrate that protein spectra present a falloff. Thus, spectral quasi-flattening may produce protein spectra with distortions when compared with experimental ones, this fact, combined with the practice of reporting only sharpened maps, generates a sub-optimal situation in terms of data preservation, reuse and reproducibility. Now that the field is more advanced, we put forward two suggestions: (1) to use methods which keep more faithfully the original experimental signal properties of macromolecules when “enhancing” the map, and (2) to further stress the need to deposit the original experimental maps without any postprocessing or sharpening, not only the enhanced maps. In the absence of access to these original maps data is lost, preventing their future analysis with new methods.