A deep learning approach to the automatic detection of alignment errors in cryo-electron tomographic reconstructions.

Electron tomography is an imaging technique that allows for the elucidation of three-dimensional structural information of biological specimens in a very general context, including cellular in situ observations. The approach starts by collecting a set of images at different projection directions by tilting the specimen stage inside the microscope. Therefore, a crucial preliminary step is to precisely define the acquisition geometry by aligning all the tilt images to a common reference. Errors introduced in this step will lead to the appearance of artifacts in the tomographic reconstruction, rendering them unsuitable for the sample study. Focusing on fiducial-based acquisition strategies, this work proposes a deep-learning algorithm to detect misalignment artifacts in tomographic reconstructions by analyzing the characteristics of these fiducial markers in the tomogram. In addition, we propose an algorithm designed to detect fiducial markers in the tomogram with which to feed the classification algorithm in case the alignment algorithm does not provide the location of the markers. This open-source software is available as part of the Xmipp software package inside of the Scipion framework, and also through the command-line in the standalone version of Xmipp.

A new algorithm for particle weighted subtraction to decrease signals from unwanted components in single particle analysis.

Single particle analysis (SPA) in cryo-electron microscopy (cryo-EM) is highly used to obtain the near-atomic structure of biological macromolecules. The current methods allow users to produce high-resolution maps from many samples. However, there are still challenging cases that require extra processing to obtain high resolution. This is the case when the macromolecule of the sample is composed of different components and we want to focus just on one of them. For example, if the macromolecule is composed of several flexible subunits and we are interested in a specific one, if it is embedded in a viral capsid environment, or if it has additional components to stabilize it, such as nanodiscs. The signal from these components, which in principle we are not interested in, can be removed from the particles using a projection subtraction method. Currently, there are two projection subtraction methods used in practice and both have some limitations. In fact, after evaluating their results, we consider that the problem is still open to new solutions, as they do not fully remove the signal of the components that are not of interest. Our aim is to develop a new and more precise projection subtraction method, improving the performance of state-of-the-art methods. We tested our algorithm with data from public databases and an in-house data set. In this work, we show that the performance of our algorithm improves the results obtained by others, including the localization of small ligands, such as drugs, whose binding location is unknown a priori.

Local defocus estimation in single particle analysis in cryo-electron microscopy.

Single Particle analysis (SPA) aims to determine the three-dimensional structure of proteins and macromolecular complexes. The current state of the art has allowed us to achieve near-atomic and even atomic resolutions. To obtain high-resolution structures, a set of well-defined image processing steps is required. A critical one is the estimation of the Contrast Transfer Function (CTF), which considers the sample defocus and aberrations of the microscope. Defocus is usually globally estimated; in this case, it is the same for all the particles in each micrograph. But proteins are ice-embedded at different heights, suggesting that defocus should be measured in a local (per particle) manner. There are four state-of-the-art programs to estimate local defocus (Gctf, Relion, CryoSPARC, and Xmipp). In this work, we have compared the results of these software packages to check whether the resolution improves. We have used the Scipion framework and developed a specific program to analyze local defocus. The results produced by different programs do not show a clear consensus using the current test datasets in this study.

Scipion Flexibility Hub: an integrative framework for advanced analysis of conformational heterogeneity in cryoEM.

Understanding how structure and function meet to drive biological processes is progressively shifting the cryoEM field towards a more advanced analysis of macromolecular flexibility. Thanks to techniques such as single-particle analysis and electron tomography, it is possible to image a macromolecule in different states, information that can subsequently be extracted through advanced image-processing methods to build a richer approximation of a conformational landscape. However, the interoperability of all of these algorithms remains a challenging task that is left to users, preventing them from defining a single flexible workflow in which conformational information can be addressed by different algorithms. Therefore, in this work, a new framework integrated into Scipion is proposed called the Flexibility Hub. This framework automatically handles intercommunication between different heterogeneity software, simplifying the task of combining the software into workflows in which the quality and the amount of information extracted from flexibility analysis is maximized.

ZART: A Novel Multiresolution Reconstruction Algorithm with Motion-blur Correction for Single Particle Analysis.

One of the main purposes of CryoEM Single Particle Analysis is to reconstruct the three-dimensional structure of a macromolecule thanks to the acquisition of many particle images representing different poses of the sample. By estimating the orientation of each projected particle, it is possible to recover the underlying 3D volume by multiple 3D reconstruction methods, usually working either in Fourier or in real space. However, the reconstruction from the projected images works under the assumption that all particles in the dataset correspond to the same conformation of the macromolecule. Although this requisite holds for some macromolecules, it is not true for flexible specimens, leading to motion-induced artefacts in the reconstructed CryoEM maps. In this work, we introduce a new Algebraic Reconstruction Technique called ZART, which is able to include continuous flexibility information during the reconstruction process to improve local resolution and reduce motion blurring. The conformational changes are modelled through Zernike3D polynomials. Our implementation allows for a multiresolution description of the macromolecule adapting itself to the local resolution of the reconstructed map. In addition, ZART has also proven to be a useful algorithm in cases where flexibility is not so dominant, as it improves the overall aspect of the reconstructed maps by improving their local and global resolution.

Estimating conformational landscapes from Cryo-EM particles by 3D Zernike polynomials.

The new developments in Cryo-EM Single Particle Analysis are helping us to understand how the macromolecular structure and function meet to drive biological processes. By capturing many states at the particle level, it is possible to address how macromolecules explore different conformations, information that is classically extracted through 3D classification. However, the limitations of classical approaches prevent us from fully understanding the complete conformational landscape due to the reduced number of discrete states accurately reconstructed. To characterize the whole structural spectrum of a macromolecule, we propose an extension of our Zernike3D approach, able to extract per-image continuous flexibility information directly from a particle dataset. Also, our method can be seamlessly applied to images, maps or atomic models, opening integrative possibilities. Furthermore, we introduce the ZART reconstruction algorithm, which considers the Zernike3D deformation fields to revert particle conformational changes during the reconstruction process, thus minimizing the blurring induced by molecular motions.

Image processing tools for the validation of CryoEM maps.

The number of maps deposited in public databases (Electron Microscopy Data Bank, EMDB) determined by cryo-electron microscopy has quickly grown in recent years. With this rapid growth, it is critical to guarantee their quality. So far, map validation has primarily focused on the agreement between maps and models. From the image processing perspective, the validation has been mostly restricted to using two half-maps and the measurement of their internal consistency. In this article, we suggest that map validation can be taken much further from the point of view of image processing if 2D classes, particles, angles, coordinates, defoci, and micrographs are also provided. We present a progressive validation scheme that qualifies a result validation status from 0 to 5 and offers three optional qualifiers (A, W, and O) that can be added. The simplest validation state is 0, while the most complete would be 5AWO. This scheme has been implemented in a website https://biocomp.cnb.csic.es/EMValidationService/ to which reconstructed maps and their ESI can be uploaded.

BIPSPI+: Mining Type-Specific Datasets of Protein Complexes to Improve Protein Binding Site Prediction.

Computational approaches for predicting protein-protein interfaces are extremely useful for understanding and modelling the quaternary structure of protein assemblies. In particular, partner-specific binding site prediction methods allow delineating the specific residues that compose the interface of protein complexes. In recent years, new machine learning and other algorithmic approaches have been proposed to solve this problem. However, little effort has been made in finding better training datasets to improve the performance of these methods. With the aim of vindicating the importance of the training set compilation procedure, in this work we present BIPSPI+, a new version of our original server trained on carefully curated datasets that outperforms our original predictor. We show how prediction performance can be improved by selecting specific datasets that better describe particular types of protein interactions and interfaces (e.g. homo/hetero). In addition, our upgraded web server offers a new set of functionalities such as the sequence-structure prediction mode, hetero- or homo-complex specialization and the guided docking tool that allows to compute 3D quaternary structure poses using the predicted interfaces. BIPSPI+ is freely available at https://bipspi.cnb.csic.es.

ScipionTomo: Towards cryo-electron tomography software integration, reproducibility, and validation.

Image processing in cryogenic electron tomography (cryoET) is currently at a similar state as Single Particle Analysis (SPA) in cryogenic electron microscopy (cryoEM) was a few years ago. Its data processing workflows are far from being well defined and the user experience is still not smooth. Moreover, file formats of different software packages and their associated metadata are not standardized, mainly since different packages are developed by different groups, focusing on different steps of the data processing pipeline. The Scipion framework, originally developed for SPA (de la Rosa-Trevín et al., 2016), has a generic python workflow engine that gives it the versatility to be extended to other fields, as demonstrated for model building (Martínez et al., 2020). In this article, we provide an extension of Scipion based on a set of tomography plugins (referred to as ScipionTomo hereafter), with a similar purpose: to allow users to be focused on the data processing and analysis instead of having to deal with multiple software installation issues and the inconvenience of switching from one to another, converting metadata files, managing possible incompatibilities, scripting (writing a simple program in a language that the computer must convert to machine language each time the program is run), etcetera. Additionally, having all the software available in an integrated platform allows comparing the results of different algorithms trying to solve the same problem. In this way, the commonalities and differences between estimated parameters shed light on which results can be more trusted than others. ScipionTomo is developed by a collaborative multidisciplinary team composed of Scipion team engineers, structural biologists, and in some cases, the developers whose software packages have been integrated. It is open to anyone in the field willing to contribute to this project. The result is a framework extension that combines the acquired knowledge of Scipion developers in close collaboration with third-party developers, and the on-demand design of functionalities requested by beta testers applying this solution to actual biological problems.

Cryo-Electron Microscopy: The field of 1,000+ methods.

Cryo-Electron Microscopy (CryoEM) is currently a well-established method to elucidate a biological macromolecule's three-dimensional (3D) structure. Its success is due to technological and methodological advances in several fronts: sample preparation, electron optics and detection, image acquisition, image processing, and map interpretation. The first methods started in the late 1960s and, since then, new methods on all fronts have continuously been published, maturating the field as we know it now. In terms of publications, we can distinguish several periods, witnessing a substantial acceleration of methodological publications in recent years, pointing out to an increased interest in the domain. On the other hand, this accelerated increase of methods development may confuse practitioners about which method they should be using (and how) and highlight the importance of paying attention to establishing best practices for methods reporting and usage. In this paper, we analyze the trends identified in over 1,000 methodological papers. Our focus is primarily on computational image processing methods. However, our list also covers some aspects of sample preparation and image acquisition. Several interesting ideas stem out from this study: (1) Single Particle Analysis (SPA) has largely accelerated in the last decade and sample preparation methods in the last five years; (2) Electron Tomography is not yet in a rapidly growing phase, but it is foreseeable that it will soon be; (3) the work horses of SPA are 3D classification, 3D reconstruction, and 3D alignment, and there have been many papers on these topics, which are not considered to be solved yet, but ever improving; and (4) since the resolution revolution, atomic modelling has also caught on as a hot topic.

On bias, variance, overfitting, gold standard and consensus in single-particle analysis by cryo-electron microscopy.

Cryo-electron microscopy (cryoEM) has become a well established technique to elucidate the 3D structures of biological macromolecules. Projection images from thousands of macromolecules that are assumed to be structurally identical are combined into a single 3D map representing the Coulomb potential of the macromolecule under study. This article discusses possible caveats along the image-processing path and how to avoid them to obtain a reliable 3D structure. Some of these problems are very well known in the community. These may be referred to as sample-related (such as specimen denaturation at interfaces or non-uniform projection geometry leading to underrepresented projection directions). The rest are related to the algorithms used. While some have been discussed in depth in the literature, such as the use of an incorrect initial volume, others have received much less attention. However, they are fundamental in any data-analysis approach. Chiefly among them, instabilities in estimating many of the key parameters that are required for a correct 3D reconstruction that occur all along the processing workflow are referred to, which may significantly affect the reliability of the whole process. In the field, the term overfitting has been coined to refer to some particular kinds of artifacts. It is argued that overfitting is a statistical bias in key parameter-estimation steps in the 3D reconstruction process, including intrinsic algorithmic bias. It is also shown that common tools (Fourier shell correlation) and strategies (gold standard) that are normally used to detect or prevent overfitting do not fully protect against it. Alternatively, it is proposed that detecting the bias that leads to overfitting is much easier when addressed at the level of parameter estimation, rather than detecting it once the particle images have been combined into a 3D map. Comparing the results from multiple algorithms (or at least, independent executions of the same algorithm) can detect parameter bias. These multiple executions could then be averaged to give a lower variance estimate of the underlying parameters.

Cryo-EM density maps adjustment for subtraction, consensus and sharpening.

Electron cryomicroscopy (cryo-EM) has emerged as a powerful structural biology instrument to solve near-atomic three-dimensional structures. Despite the fast growth in the number of density maps generated from cryo-EM data, comparison tools among these reconstructions are still lacking. Current proposals to compare cryo-EM data derived volumes perform map subtraction based on adjustment of each volume grey level to the same scale. We present here a more sophisticated way of adjusting the volumes before comparing, which implies adjustment of grey level scale and spectrum energy, but keeping phases intact inside a mask and imposing the results to be strictly positive. The adjustment that we propose leaves the volumes in the same numeric frame, allowing to perform operations among the adjusted volumes in a more reliable way. This adjustment can be a preliminary step for several applications such as comparison through subtraction, map sharpening, or combination of volumes through a consensus that selects the best resolved parts of each input map. Our development might also be used as a sharpening method using an atomic model as a reference. We illustrate the applicability of this algorithm with the reconstructions derived of several experimental examples. This algorithm is implemented in Xmipp software package and its applications are user-friendly accessible through the cryo-EM image processing framework Scipion.

Cryo-EM and Single-Particle Analysis with Scipion.

Cryo-electron microscopy has become one of the most important tools in biological research to reveal the structural information of macromolecules at near-atomic resolution. In single-particle analysis, the vitrified sample is imaged by an electron beam and the detectors at the end of the microscope column produce movies of that sample. These movies contain thousands of images of identical particles in random orientations. The data need to go through an image processing workflow with multiple steps to obtain the final 3D reconstructed volume. The goal of the image processing workflow is to identify the acquisition parameters to be able to reconstruct the specimen under study. Scipion provides all the tools to create this workflow using several image processing packages in an integrative framework, also allowing the traceability of the results. In this article the whole image processing workflow in Scipion is presented and discussed with data coming from a real test case, giving all the details necessary to go from the movies obtained by the microscope to a high resolution final 3D reconstruction. Also, the power of using consensus tools that allow combining methods, and confirming results along every step of the workflow, improving the accuracy of the obtained results, is discussed.

DeepAlign, a 3D alignment method based on regionalized deep learning for Cryo-EM.

Cryo Electron Microscopy (Cryo-EM) is currently one of the main tools to reveal the structural information of biological specimens at high resolution. Despite the great development of the techniques involved to solve the biological structures with Cryo-EM in the last years, the reconstructed 3D maps can present lower resolution due to errors committed while processing the information acquired by the microscope. One of the main problems comes from the 3D alignment step, which is an error-prone part of the reconstruction workflow due to the very low signal-to-noise ratio (SNR) common in Cryo-EM imaging. In fact, as we will show in this work, it is not unusual to find a disagreement in the alignment parameters in approximately 20-40% of the processed images, when outputs of different alignment algorithms are compared. In this work, we present a novel method to align sets of single particle images in the 3D space, called DeepAlign. Our proposal is based on deep learning networks that have been successfully used in plenty of problems in image classification. Specifically, we propose to design several deep neural networks on a regionalized basis to classify the particle images in sub-regions and, then, make a refinement of the 3D alignment parameters only inside that sub-region. We show that this method results in accurately aligned images, improving the Fourier shell correlation (FSC) resolution obtained with other state-of-the-art methods while decreasing computational time.

ENRICH: A fast method to improve the quality of flexible macromolecular reconstructions.

Cryo-electron microscopy using single particle analysis requires the computational averaging of thousands of projection images captured from identical macromolecules. However, macromolecules usually present some degree of flexibility showing different conformations. Computational approaches are then required to classify heterogeneous single particle images into homogeneous sets corresponding to different structural states. Nonetheless, sometimes the attainable resolution of reconstructions obtained from these smaller homogeneous sets is compromised because of reduced number of particles or lack of images at certain macromolecular orientations. In these situations, the current solution to improve map resolution is returning to the electron microscope and collect more data. In this work, we present a fast approach to partially overcome this limitation for heterogeneous data sets. Our method is based on deforming and then moving particles between different conformations using an optical flow approach. Particles are then merged into a unique conformation obtaining reconstructions with improved resolution, contrast and signal-to-noise ratio. We present experimental results that show clear improvements in the quality of obtained 3D maps, however, there are also limits to this approach, i.e., the method is restricted to small deformations and cannot determine local patterns of flexibility of small elements, such as secondary structures, which we discuss in the manuscript.

Algorithmic robustness to preferred orientations in single particle analysis by CryoEM.

The presence of preferred orientations in single particle analysis (SPA) by cryo-Electron Microscopy (cryoEM) is currently one of the hurdles preventing many structural analyses from yielding high-resolution structures. Although the existence of preferred orientations is mostly related to the grid preparation, in this technical note, we show that some image processing algorithms used for angular assignment and three-dimensional (3D) reconstruction are more robust than others to these detrimental conditions. We exemplify this argument with three different data sets in which the presence of preferred orientations hindered achieving a 3D reconstruction without artifacts or, even worse, a 3D reconstruction could never be achieved.

Improvements on marker-free images alignment for electron tomography.

Electron tomography is a technique to obtain three-dimensional structural information of samples. However, the technique is limited by shifts occurring during acquisition that need to be corrected before the reconstruction process. In 2009, we proposed an approach for post-acquisition alignment of tilt series images. This approach was marker-free, based on patch tracking and integrated in free software. Here, we present improvements to the method to make it more reliable, stable and accurate. In addition, we modified the image formation model underlying the alignment procedure to include different deformations occurring during acquisition. We propose a new way to correct these computed deformations to obtain reconstructions with reduced artifacts. The new approach has demonstrated to improve the quality of the final 3D reconstruction, giving access to better defined structures for different transmission electron tomography methods: resin embedded STEM-tomography and cryo-TEM tomography. The method is freely available in TomoJ software.

Measurement of local resolution in electron tomography.

Resolution (global and local) is one of the most reported metrics of quality measurement in Single Particle Analysis (SPA). However, in electron tomography, the situation is different and its computation is not straightforward. Typically, resolution estimation is global and, therefore, reduces the assessment of a whole tomogram to a single number. However, it is known that tomogram quality is spatially variant. Still, up to our knowledge, a method to estimate local quality metrics in tomography is lacking. This work introduces MonoTomo, a method developed to estimate locally in a tomogram the highest reliable frequency component, expressed as a form of local resolution. The fundamentals lie in a local analysis of the density map via monogenic signals, which, in analogy to MonoRes, allows for local estimations. Results with experimental data show that the local resolution range that MonoTomo casts agrees with reported resolution values for experimental data sets, with the advantage of providing a local estimation. A range of applications of MonoTomo are suggested for further exploration.

Local resolution estimates of cryoEM reconstructions.

The field of cryoEM has quickly advanced in last years with the new biochemical, technological, methodological and computational developments. It has allowed significant progresses in Structural Biology, typically reaching quasi-atomic resolutions in the reconstructed maps. However, this rapid advance has also generated new questions relevant to resolution estimates. The global resolution metrics and their criteria have been deeply discussed in the last decade, but despite that, it remains as an important issue in the field. Recently, the introduction of local resolution measurements has changed how cryoEM reconstructions are interpreted, providing information about the existence of heterogeneity, flexibility, and angular assignment errors, and using it as a tool to aid in modeling. In this review we revisit the concept of local resolution and the different algorithms in the current state of the art. However, the concept of local resolution is not uniquely defined, and each implementation measures different features. This may lead to inappropriate interpretation of local resolution maps. Hence, a set of good practices is provided in this review to avoid misleading and over-interpretation of the reconstructions.

MicrographCleaner: A python package for cryo-EM micrograph cleaning using deep learning.

Cryo-EM Single Particle Analysis workflows require tens of thousands of high-quality particle projections to unveil the three-dimensional structure of macromolecules. Conventional methods for automatic particle picking tend to suffer from high false-positive rates, hampering the reconstruction process. One common cause of this problem is the presence of carbon and different types of high-contrast contaminations. In order to overcome this limitation, we have developed MicrographCleaner, a deep learning package designed to discriminate, in an automated fashion, between regions of micrographs which are suitable for particle picking, and those which are not. MicrographCleaner implements a U-net-like deep learning model trained on a manually curated dataset compiled from over five hundred micrographs. The benchmarking, carried out on approximately one hundred independent micrographs, shows that MicrographCleaner is a very efficient approach for micrograph preprocessing. MicrographCleaner (micrograph_cleaner_em) package is available at PyPI and Anaconda Cloud and also as a Scipion/Xmipp protocol. Source code is available at https://github.com/rsanchezgarc/micrograph_cleaner_em.