Computational methods for constructing protein structure models from 3D electron microscopy maps

Artificial Intelligence in Cryo-Electron Microscopy

Cryo-electron microscopy (cryo-EM) has become an unrivaled tool for determining the structure of macromolecular complexes. The biological function of macromolecular complexes is inextricably tied to the flexibility of these complexes. Single particle cryo-EM can reveal the conformational heterogeneity of a biochemically pure sample, leading to well-founded mechanistic hypotheses about the roles these complexes play in biology. However, the processing of increasingly large, complex datasets using traditional data processing strategies is exceedingly expensive in both user time and computational resources. Current innovations in data processing capitalize on artificial intelligence (AI) to improve the efficiency of data analysis and validation. Here, we review new tools that use AI to automate the data analysis steps of particle picking, 3D map reconstruction, and local resolution determination. We discuss how the application of AI moves the field forward, and what obstacles remain. We also introduce potential future applications of AI to use cryo-EM in understanding protein communities in cells.

MarkovFit: Structure Fitting for Protein Complexes in Electron Microscopy Maps Using Markov Random Field

An increasing number of protein complex structures are determined by cryo-electron microscopy (cryo-EM). When individual protein structures have been determined and are available, an important task in structure modeling is to fit the individual structures into the density map. Here, we designed a method that fits the atomic structures of proteins in cryo-EM maps of medium to low resolutions using Markov random fields, which allows probabilistic evaluation of fitted models. The accuracy of our method, MarkovFit, performed better than existing methods on datasets of 31 simulated cryo-EM maps of resolution 10 Å , nine experimentally determined cryo-EM maps of resolution less than 4 Å , and 28 experimentally determined cryo-EM maps of resolution 6 to 20 Å .

Protein Structural Modeling for Electron Microscopy Maps Using VESPER and MAINMAST

An increasing number of protein structures are determined by cryo-electron microscopy (cryo-EM) and stored in the Electron Microscopy Data Bank (EMDB). To interpret determined cryo-EM maps, several methods have been developed that model the tertiary structure of biomolecules, particularly proteins. Here we show how to use two such methods, VESPER and MAINMAST, which were developed in our group. VESPER is a method mainly for two purposes: fitting protein structure models into an EM map and aligning two EM maps locally or globally to capture their similarity. VESPER represents each EM map as a set of vectors pointing toward denser points. By considering matching the directions of vectors, in general, VESPER aligns maps better than conventional methods that only consider local densities of maps. MAINMAST is a de novo protein modeling tool designed for EM maps with resolution of 3-5 Å or better. MAINMAST builds a protein main chain directly from a density map by tracing dense points in an EM map and connecting them using a tree-graph structure. This article describes how to use these two tools using three illustrative modeling examples. © 2022 The Authors. Current Protocols published by Wiley Periodicals LLC. Basic Protocol 1: Protein structure model fitting using VESPER Alternate Protocol: Atomic model fitting using VESPER web server Basic Protocol 2: Protein de novo modeling using MAINMAST.

Interface refinement of low- to medium-resolution Cryo-EM complexes using HADDOCK2.4

A wide range of cellular processes requires the formation of multimeric protein complexes. The rise of cryo-electron microscopy (cryo-EM) has enabled the structural characterization of these protein assemblies. The density maps produced can, however, still suffer from limited resolution, impeding the process of resolving structures at atomic resolution. In order to solve this issue, monomers can be fitted into low- to medium-resolution maps. Unfortunately, the models produced frequently contain atomic clashes at the protein-protein interfaces (PPIs), as intermolecular interactions are typically not considered during monomer fitting. Here, we present a refinement approach based on HADDOCK2.4 to remove intermolecular clashes and optimize PPIs. A dataset of 14 cryo-EM complexes was used to test eight protocols. The best-performing protocol, consisting of a semi-flexible simulated annealing refinement with centroid restraints on the monomers, was able to decrease intermolecular atomic clashes by 98% without significantly deteriorating the quality of the cryo-EM density fit.

Deep Learning-Based Advances in Protein Structure Prediction

Obtaining an accurate description of protein structure is a fundamental step toward understanding the underpinning of biology. Although recent advances in experimental approaches have greatly enhanced our capabilities to experimentally determine protein structures, the gap between the number of protein sequences and known protein structures is ever increasing. Computational protein structure prediction is one of the ways to fill this gap. Recently, the protein structure prediction field has witnessed a lot of advances due to Deep Learning (DL)-based approaches as evidenced by the success of AlphaFold2 in the most recent Critical Assessment of protein Structure Prediction (CASP14). In this article, we highlight important milestones and progresses in the field of protein structure prediction due to DL-based methods as observed in CASP experiments. We describe advances in various steps of protein structure prediction pipeline viz. protein contact map prediction, protein distogram prediction, protein real-valued distance prediction, and Quality Assessment/refinement. We also highlight some end-to-end DL-based approaches for protein structure prediction approaches. Additionally, as there have been some recent DL-based advances in protein structure determination using Cryo-Electron (Cryo-EM) microscopy based, we also highlight some of the important progress in the field. Finally, we provide an outlook and possible future research directions for DL-based approaches in the protein structure prediction arena.

Coarse-Grained Simulation of Mechanical Properties of Single Microtubules With Micrometer Length

Microtubules are one of the most important components in the cytoskeleton and play a vital role in maintaining the shape and function of cells. Because single microtubules are some micrometers long, it is difficult to simulate such a large system using an all-atom model. In this work, we use the newly developed convolutional and K-means coarse-graining (CK-CG) method to establish an ultra-coarse-grained (UCG) model of a single microtubule, on the basis of the low electron microscopy density data of microtubules. We discuss the rationale of the micro-coarse-grained microtubule models of different resolutions and explore microtubule models up to 12-micron length. We use the devised microtubule model to quantify mechanical properties of microtubules of different lengths. Our model allows mesoscopic simulations of micrometer-level biomaterials and can be further used to study important biological processes related to microtubule function.

Homology modeling in the time of collective and artificial intelligence

Homology modeling is a method for building protein 3D structures using protein primary sequence and utilizing prior knowledge gained from structural similarities with other proteins. The homology modeling process is done in sequential steps where sequence/structure alignment is optimized, then a backbone is built and later, side-chains are added. Once the low-homology loops are modeled, the whole 3D structure is optimized and validated. In the past three decades, a few collective and collaborative initiatives allowed for continuous progress in both homology and ab initio modeling. Critical Assessment of protein Structure Prediction (CASP) is a worldwide community experiment that has historically recorded the progress in this field. Folding@Home and Rosetta@Home are examples of crowd-sourcing initiatives where the community is sharing computational resources, whereas RosettaCommons is an example of an initiative where a community is sharing a codebase for the development of computational algorithms. Foldit is another initiative where participants compete with each other in a protein folding video game to predict 3D structure. In the past few years, contact maps deep machine learning was introduced to the 3D structure prediction process, adding more information and increasing the accuracy of models significantly. In this review, we will take the reader in a journey of exploration from the beginnings to the most recent turnabouts, which have revolutionized the field of homology modeling. Moreover, we discuss the new trends emerging in this rapidly growing field.

Protein-Protein Modeling Using Cryo-EM Restraints

Recent improvements in cryo-electron microscopy (cryo-EM) in the past few years are now allowing to observe molecular complexes at atomic resolution. As a consequence, numerous structures derived from cryo-EM are now available in the Protein Data Bank. However, if for some complexes atomic resolution is reached, this is not true for all. This is also the case in cryo-electron tomography where the achievable resolution is still limited. Furthermore the resolution in a cryo-EM map is not a constant, with often outer regions being of lower resolution, possibly linked to conformational variability. Although those low- to medium-resolution EM maps (or regions thereof) cannot directly provide atomic structure of large molecular complexes, they provide valuable information to model the individual components and their assembly into them. Most approaches for this kind of modeling are performing rigid fitting of the individual components into the EM density map. While this would appear an obvious option, they ignore key aspects of molecular recognition, the energetics and flexibility of the interfaces. Moreover, this often restricts the modeling to a unique source of data, the EM density map.In this chapter, we describe a protocol where an EM map is used as restraint in HADDOCK to guide the modeling process. In the first step, rigid-body fitting is performed with PowerFit in order to identify the most likely locations of the molecules into the map. These are then used as centroids to which distance restraints are defined from the center of mass of the components of the complex for the initial rigid-body docking. The EM density is then directly used as an additional restraint energy term, which can be combined with all the other types of data supported by HADDOCK. This protocol relies on the new version 2.4 of both the HADDOCK webserver and software. Preparation steps consisting of cropping the EM map and rigid-body fitting of the atomic structure are explained. Then, the EM-driven docking protocol using HADDOCK is illustrated.

Protein secondary structure detection in intermediate-resolution cryo-EM maps using deep learning

An increasing number of protein structures have been solved by cryo-electron microscopy (cryo-EM). Although structures determined at near-atomic resolution are now routinely reported, many density maps are still determined at an intermediate resolution, where extracting structure information is still a challenge. We have developed a computational method, Emap2sec, which identifies the secondary structures of proteins (α helices, β sheets, and other structures) in an EM map of 5 to 10 Å resolution. Emap2sec uses a 3D deep convolutional neural network to assign secondary structure to each grid point in an EM map. We tested Emap2sec on 6.0 and 10.0 Å resolution EM maps simulated from 34 structures, as well as on 43 maps determined experimentally at 5.0 to 9.5 Å resolution. Emap2sec was able to clearly identify the secondary structures in many maps tested, and showed substantially better performance than existing methods.

Advances in image processing for single-particle analysis by electron cryomicroscopy and challenges ahead

Electron cryomicroscopy (cryoEM) is essential for the study and functional understanding of non-crystalline macromolecules such as proteins. These molecules cannot be imaged using X-ray crystallography or other popular methods. CryoEM has been successfully used to visualize macromolecular complexes such as ribosomes, viruses, and ion channels. Determination of structural models of these at various conformational states leads to insight on how these molecules function. Recent advances in imaging technology have given cryoEM a scientific rebirth. As a result of these technological advances image processing and analysis have yielded molecular structures at atomic resolution. Nevertheless there continue to be challenges in image processing, and in this article we will touch on the most essential in order to derive an accurate three-dimensional model from noisy projection images. Traditional approaches, such as k-means clustering for class averaging, will be provided as background. We will then highlight new approaches for each image processing subproblem, including a 3D reconstruction method for asymmetric molecules using just two projection images and deep learning algorithms for automated particle picking.

De novo main-chain modeling with MAINMAST in 2015/2016 EM Model Challenge

Protein tertiary structure modeling is a critical step for the interpretation of three dimensional (3D) election microscopy density. Our group participated the 2015/2016 EM Model Challenge using the MAINMAST software for a de novo main chain modeling. The software generates local dense points using the mean shifting algorithm, and connects them into Cα models by calculating the minimum spanning tree and the longest path. Subsequently, full atom structure models are generated, which are subject to structural refinement. Here, we summarize the qualities of our submitted models and examine successful and unsuccessful models, including 3D models we did not submit to the Challenge. Our protocol using the MAINMAST software was sometimes able to build correct conformations with 3.4–5.1 Å RMSD. Unsuccessful models had failure of chain traces, however, their Cα positions and some local structures were quite correctly built. For evaluate the quality of the models, the MAINMAST software provides a confidence score for each Cα position from the consensus of top 100 scoring models.

Advances in the Development of Shape Similarity Methods and Their Application in Drug Discovery

Molecular similarity is a key concept in drug discovery. It is based on the assumption that structurally similar molecules frequently have similar properties. Assessment of similarity between small molecules has been highly effective in the discovery and development of various drugs. Especially, two-dimensional (2D) similarity approaches have been quite popular due to their simplicity, accuracy and efficiency. Recently, the focus has been shifted toward the development of methods involving the representation and comparison of three-dimensional (3D) conformation of small molecules. Among the 3D similarity methods, evaluation of shape similarity is now gaining attention for its application not only in virtual screening but also in molecular target prediction, drug repurposing and scaffold hopping. A wide range of methods have been developed to describe molecular shape and to determine the shape similarity between small molecules. The most widely used methods include atom distance-based methods, surface-based approaches such as spherical harmonics and 3D Zernike descriptors, atom-centered Gaussian overlay based representations. Several of these methods demonstrated excellent virtual screening performance not only retrospectively but also prospectively. In addition to methods assessing the similarity between small molecules, shape similarity approaches have been developed to compare shapes of protein structures and binding pockets. Additionally, shape comparisons between atomic models and 3D density maps allowed the fitting of atomic models into cryo-electron microscopy maps. This review aims to summarize the methodological advances in shape similarity assessment highlighting advantages, disadvantages and their application in drug discovery.

De novo main-chain modeling for EM maps using MAINMAST

An increasing number of protein structures are determined by cryo-electron microscopy (cryo-EM) at near atomic resolution. However, tracing the main-chains and building full-atom models from EM maps of ~4-5 Å is still not trivial and remains a time-consuming task. Here, we introduce a fully automated de novo structure modeling method, MAINMAST, which builds three-dimensional models of a protein from a near-atomic resolution EM map. The method directly traces the protein's main-chain and identifies Cα positions as tree-graph structures in the EM map. MAINMAST performs significantly better than existing software in building global protein structure models on data sets of 40 simulated density maps at 5 Å resolution and 30 experimentally determined maps at 2.6-4.8 Å resolution. In another benchmark of building missing fragments in protein models for EM maps, MAINMAST builds fragments of 11-161 residues long with an average RMSD of 2.68 Å.

Bayesian Modeling of Biomolecular Assemblies with Cryo-EM Maps

A growing array of experimental techniques allows us to characterize the three-dimensional structure of large biological assemblies at increasingly higher resolution. In addition to X-ray crystallography and nuclear magnetic resonance in solution, new structure determination methods such cryo-electron microscopy (cryo-EM), crosslinking/mass spectrometry and solid-state NMR have emerged. Often it is not sufficient to use a single experimental method, but complementary data need to be collected by using multiple techniques. The integration of all datasets can only be achieved by computational means. This article describes Inferential structure determination, a Bayesian approach to integrative modeling of biomolecular complexes with hybrid structural data. I will introduce probabilistic models for cryo-EM maps and outline Markov chain Monte Carlo algorithms for sampling model structures from the posterior distribution. I will focus on rigid and flexible modeling with cryo-EM data and discuss some of the computational challenges of Bayesian inference in the context of biomolecular modeling.

Variability of Protein Structure Models from Electron Microscopy

An increasing number of biomolecular structures are solved by electron microscopy (EM). However, the quality of structure models determined from EM maps vary substantially. To understand to what extent structure models are supported by information embedded in EM maps, we used two computational structure refinement methods to examine how much structures can be refined using a dataset of 49 maps with accompanying structure models. The extent of structure modification as well as the disagreement between refinement models produced by the two computational methods scaled inversely with the global and the local map resolutions. A general quantitative estimation of deviations of structures for particular map resolutions are provided. Our results indicate that the observed discrepancy between the deposited map and the refined models is due to the lack of structural information present in EM maps and thus these annotations must be used with caution for further applications.

Use of evolutionary information in the fitting of atomic level protein models in low resolution cryo-EM map of a protein assembly improves the accuracy of the fitting

Protein-protein interface residues, especially those at the core of the interface, exhibit higher conservation than residues in solvent exposed regions. Here, we explore the ability of this differential conservation to evaluate fittings of atomic models in low-resolution cryo-EM maps and select models from the ensemble of solutions that are often proposed by different model fitting techniques. As a prelude, using a non-redundant and high-resolution structural dataset involving 125 permanent and 95 transient complexes, we confirm that core interface residues are conserved significantly better than nearby non-interface residues and this result is used in the cryo-EM map analysis. From the analysis of inter-component interfaces in a set of fitted models associated with low-resolution cryo-EM maps of ribosomes, chaperones and proteasomes we note that a few poorly conserved residues occur at interfaces. Interestingly a few conserved residues are not in the interface, though they are close to the interface. These observations raise the potential requirement of refitting the models in the cryo-EM maps. We show that sampling an ensemble of models and selection of models with high residue conservation at the interface and in good agreement with the density helps in improving the accuracy of the fit. This study indicates that evolutionary information can serve as an additional input to improve and validate fitting of atomic models in cryo-EM density maps.

De Novo modeling in cryo-EM density maps with Pathwalking

As electron cryo-microscopy (cryo-EM) can now frequently achieve near atomic resolution, accurate interpretation of these density maps in terms of atomistic detail has become paramount in deciphering macromolecular structure and function. However, there are few software tools for modeling protein structure from cryo-EM density maps in this resolution range. Here, we present an extension of our original Pathwalking protocol, which can automatically trace a protein backbone directly from a near-atomic resolution (3-6Å) density map. The original Pathwalking approach utilized a Traveling Salesman Problem solver for backbone tracing, but manual adjustment was still required during modeling. In the new version, human intervention is minimized and we provide a more robust approach for backbone modeling. This includes iterative secondary structure identification, termini detection and the ability to model multiple subunits without prior segmentation. Overall, the new Pathwalking procedure provides a more complete and robust tool for annotating protein structure function in near-atomic resolution density maps.

Improved cryoEM-Guided Iterative Molecular Dynamics--Rosetta Protein Structure Refinement Protocol for High Precision Protein Structure Prediction

Many excellent methods exist that incorporate cryo-electron microscopy (cryoEM) data to constrain computational protein structure prediction and refinement. Previously, it was shown that iteration of two such orthogonal sampling and scoring methods – Rosetta and molecular dynamics (MD) simulations – facilitated exploration of conformational space in principle. Here, we go beyond a proof-of-concept study and address significant remaining limitations of the iterative MD–Rosetta protein structure refinement protocol. Specifically, all parts of the iterative refinement protocol are now guided by medium-resolution cryoEM density maps, and previous knowledge about the native structure of the protein is no longer necessary. Models are identified solely based on score or simulation time. All four benchmark proteins showed substantial improvement through three rounds of the iterative refinement protocol. The best-scoring final models of two proteins had sub-Ångstrom RMSD to the native structure over residues in secondary structure elements. Molecular dynamics was most efficient in refining secondary structure elements and was thus highly complementary to the Rosetta refinement which is most powerful in refining side chains and loop regions.

γ-TEMPy: Simultaneous Fitting of Components in 3D-EM Maps of Their Assembly Using a Genetic Algorithm

We have developed a genetic algorithm for building macromolecular complexes using only a 3D-electron microscopy density map and the atomic structures of the relevant components. For efficient sampling the method uses map feature points calculated by vector quantization. The fitness function combines a mutual information score that quantifies the goodness of fit with a penalty score that helps to avoid clashes between components. Testing the method on ten assemblies (containing 3–8 protein components) and simulated density maps at 10, 15, and 20 Å resolution resulted in identification of the correct topology in 90%, 70%, and 60% of the cases, respectively. We further tested it on four assemblies with experimental maps at 7.2–23.5 Å resolution, showing the ability of the method to identify the correct topology in all cases. We have also demonstrated the importance of the map feature-point quality on assembly fitting in the lack of additional experimental information.

•

γ-TEMPy uses a genetic algorithm to fit multiple components into 3D-EM density maps

•

The fitness score is a combination of a Mutual Information score and a clash penalty

•

Efficient sampling is aided by using map feature points from vector quantization

•

Native topologies for assemblies containing up to eight components can be predicted

Electron Microscopy and Image Processing: Essential Tools for Structural Analysis of Macromolecules

Macromolecular electron microscopy typically depicts the structures of macromolecular complexes ranging from ∼200 kDa to hundreds of MDa. The amount of specimen required, a few micrograms, is typically 100 to 1000 times less than needed for X-ray crystallography or nuclear magnetic resonance spectroscopy. Micrographs of frozen-hydrated (cryogenic) specimens portray native structures, but the original images are noisy. Computational averaging reduces noise, and three-dimensional reconstructions are calculated by combining different views of free-standing particles ("single-particle analysis"). Electron crystallography is used to characterize two-dimensional arrays of membrane proteins and very small three-dimensional crystals. Under favorable circumstances, near-atomic resolutions are achieved. For structures at somewhat lower resolution, pseudo-atomic models are obtained by fitting high-resolution components into the density. Time-resolved experiments describe dynamic processes. Electron tomography allows reconstruction of pleiomorphic complexes and subcellular structures and modeling of macromolecules in their cellular context. Significant information is also obtained from metal-coated and dehydrated specimens.

TEMPy: a Python library for assessment of three-dimensional electron microscopy density fits

TEMPy is an object-oriented Python library that provides the means to validate density fits in electron microscopy reconstructions. This article highlights several features of particular interest for this purpose and includes some customized examples.

Three-dimensional electron microscopy is currently one of the most promising techniques used to study macromolecular assemblies. Rigid and flexible fitting of atomic models into density maps is often essential to gain further insights into the assemblies they represent. Currently, tools that facilitate the assessment of fitted atomic models and maps are needed. TEMPy (template and electron microscopy comparison using Python) is a toolkit designed for this purpose. The library includes a set of methods to assess density fits in intermediate-to-low resolution maps, both globally and locally. It also provides procedures for single-fit assessment, ensemble generation of fits, clustering, and multiple and consensus scoring, as well as plots and output files for visualization purposes to help the user in analysing rigid and flexible fits. The modular nature of TEMPy helps the integration of scoring and assessment of fits into large pipelines, making it a tool suitable for both novice and expert structural biologists.

High-resolution modeling of protein structures based on flexible fitting of low-resolution structural data

To circumvent the difficulty of directly solving high-resolution biomolecular structures, low-resolution structural data from Cryo-electron microscopy (EM) and small angle solution X-ray scattering (SAXS) are increasingly used to explore multiple conformational states of biomolecular assemblies. One promising venue to obtain high-resolution structural models from low-resolution data is via data-constrained flexible fitting. To this end, we have developed a new method based on a coarse-grained Cα-only protein representation, and a modified form of the elastic network model (ENM) that allows large-scale conformational changes while maintaining the integrity of local structures including pseudo-bonds and secondary structures. Our method minimizes a pseudo-energy which linearly combines various terms of the modified ENM energy with an EM/SAXS-fitting score and a collision energy that penalizes steric collisions. Unlike some previous flexible fitting efforts using the lowest few normal modes, our method effectively utilizes all normal modes so that both global and local structural changes can be fully modeled with accuracy. This method is also highly efficient in computing time. We have demonstrated our method using adenylate kinase as a test case which undergoes a large open-to-close conformational change. The EM-fitting method is available at a web server (http://enm.lobos.nih.gov), and the SAXS-fitting method is available as a pre-compiled executable upon request.

Conformational States of macromolecular assemblies explored by integrative structure calculation

A detailed description of macromolecular assemblies in multiple conformational states can be very valuable for understanding cellular processes. At present, structural determination of most assemblies in different biologically relevant conformations cannot be achieved by a single technique and thus requires an integrative approach that combines information from multiple sources. Different techniques require different computational methods to allow efficient and accurate data processing and analysis. Here, we summarize the latest advances and future challenges in computational methods that help the interpretation of data from two techniques—mass spectrometry and three-dimensional cryo-electron microscopy (with focus on alignment and classification of heterogeneous subtomograms from cryo-electron tomography). We evaluate how new developments in these two broad fields will lead to further integration with atomic structures to broaden our picture of the dynamic behavior of assemblies in their native environment.

Combined approaches to flexible fitting and assessment in virus capsids undergoing conformational change

Fitting of atomic components into electron cryo-microscopy (cryoEM) density maps is routinely used to understand the structure and function of macromolecular machines. Many fitting methods have been developed, but a standard protocol for successful fitting and assessment of fitted models has yet to be agreed upon among the experts in the field. Here, we created and tested a protocol that highlights important issues related to homology modelling, density map segmentation, rigid and flexible fitting, as well as the assessment of fits. As part of it, we use two different flexible fitting methods (Flex-EM and iMODfit) and demonstrate how combining the analysis of multiple fits and model assessment could result in an improved model. The protocol is applied to the case of the mature and empty capsids of Coxsackievirus A7 (CAV7) by flexibly fitting homology models into the corresponding cryoEM density maps at 8.2 and 6.1 Å resolution. As a result, and due to the improved homology models (derived from recently solved crystal structures of a close homolog – EV71 capsid – in mature and empty forms), the final models present an improvement over previously published models. In close agreement with the capsid expansion observed in the EV71 structures, the new CAV7 models reveal that the expansion is accompanied by ∼5° counterclockwise rotation of the asymmetric unit, predominantly contributed by the capsid protein VP1. The protocol could be applied not only to viral capsids but also to many other complexes characterised by a combination of atomic structure modelling and cryoEM density fitting.