AI-based Algorithms for Cryo-Electron Microscopy: Comparison
Please note this is a comparison between Version 3 by Jessie Wu and Version 2 by Jessie Wu.

Single particle cryo-electron microscopy (cryo-EM) has emerged as the prevailing method for near-atomic structure determination, shedding light on the important molecular mechanisms of biological macromolecules. However, the inherent dynamics and structural variability of biological complexes coupled with the large number of experimental images generated by a cryo-EM experiment make data processing nontrivial. In particular, ab initio reconstruction and atomic model building remain major bottlenecks that demand substantial computational resources and manual intervention. Approaches utilizing recent innovations in artificial intelligence (AI) technology, particularly deep learning, have the potential to overcome the limitations that cannot be adequately addressed by traditional image processing approaches.

  • cryo-electron microscopy
  • cryo-EM
  • deep learning
  • machine learning
  • artificial intelligence
  • AI
  • neural networks

1. Introduction

Cryo-electron microscopy (Cryo-EM) and single particle analysis (SPA) have become the preferred method for structure determination of biological complexes at near-atomic resolution [1][2][3][4][5][6]. Such level of detail not only provides crucial insight into molecular mechanisms employed by biological macromolecules, but also may facilitate the design and development of new drugs and therapeutics. The widespread adoption of cryo-EM is primarily credited to recent technological advances, including the introduction of direct electron detectors and improvements in computer hardware and image-processing software [7] that have enabled routine structure determination at high-resolution (reviewed in [8]). In contrast to traditional structural biology techniques, such as X-ray crystallography and Nuclear Magnetic Resonance (NMR), cryo-EM requires very small amounts of purified protein in the range of 0.1 to 5 mg/mL [9]. Additionally, SPA has proven to be a powerful tool for determining the structures of large and dynamic biological complexes that exhibit a range of compositional and conformational heterogeneity [10][11][12].
The goal of SPA is to calculate a high-resolution 3D structure from the noisy, 2D projections of the specimen produced on the direct electron detector by the beam of high energy electrons. Multiple software packages have been developed to facilitate this process [13][14][15][16][17][18][19][20], and a typical SPA workflow is described in Figure 1. However, several factors complicate image processing: (1) due to advancements in microscope optics, detector technology, and data storage hardware, a single cryo-EM experiment can generate thousands of micrographs with millions of particle images [21][22]; (2) biological assemblies are often dynamic and flexible molecular machines that adopt a variety of structural states; (3) cryo-EM micrographs have a low signal-to-noise ratio (SNR), and they are susceptible to artifacts, such as ice contaminations, radiation damage, and preferred orientation of imaged particles, which can obscure underlying structural information. While the application of traditional machine learning (ML) techniques has already enabled the automation of many stages of the workflow [23][24][25][26][27][28], ab initio reconstruction and atomic model building remain major hurdles in need of robust methodologies to minimize the extensive computational resources and manual user input they currently demand.
Figure 1. A typical SPA workflow. Movies collected with an electron microscope are first motion-corrected. In this step, frames are aligned and averaged to account for beam-induced motion, which increases the SNR of images. The resultant micrographs undergo contrast transfer function (CTF) estimation to calculate the effects of defocus and microscope aberrations. This step is followed by particle picking and extraction, in which particles are selected and extracted from micrographs. The extracted particles are sorted based on orientation into discrete, 2D classes, and the user removes classes containing non-particles, noise, and artifacts. Such filtered particle stacks are used to generate one or more low-resolution ab initio reconstructions that are iteratively refined through 3D classification and 3D refinement to yield final Coulomb potential maps. Given sufficient map resolution and quality, atomic models can be built and validated.

2. Artificial Intelligence-Based Approaches to De Novo Model Building

Early applications of AI methods to model building utilized conventional ML techniques such as k-nearest neighbor (k-NN) [29], k-means clustering [30], and support-vector machines (SVM) [31]. These approaches have been successful in identifying SSEs [29][31] and modeling simplified backbone structures [30]. For example, the program RENNSH [29] identifies α-helices in EM maps by representing each voxel as a spherical harmonic descriptor and using a nested k-NN framework to classify α-helix voxels. Another ML approach, SSELearner [31], first learns from maps deposited in the EMDB and then applies an SVM classifier to detect both α-helices and β-sheets in intermediate-resolution density maps. Pathwalking [30] is a de novo model-building approach that uses a combination of the traveling salesman problem and k-means clustering to build a Cα model. Nonetheless, these programs do not guarantee convergence to the correct solution and are not capable of building complete atomic models [32][33].
DL has recently driven remarkable advancements in structural biology, including the development of programs for automated de novo atomic model building from cryo-EM density maps (Table 1). This breakthrough is largely attributed to increased computational power, quality of EM images, and the number of available high-resolution cryo-EM structures for model training [33]. Recently, many CNN-based approaches have been applied not only for automatic assignment of SSEs [34][35][36][37] and backbone chains [37][38][39][40], but also to detect individual amino acids [40][41], thus generating complete de novo models [37][38][39][40][41] (Table 1). Below, wresearchers highlight some representative programs utilizing CNN architectures, followed by a description of methods that apply alternative DL frameworks to model building.
AAnchor [40] is the first program with 3D CNN architecture to identify individual amino acids in EM maps. This technique utilizes a classification CNN to locate and label residues, known as anchors, with highest confidence within a defined voxel size [40]. However, AAnchor is currently limited to maps with 3.1 Å resolution or higher and on average, the fraction of anchors detected ranges from 10 to 20% depending on map resolution [40]. Another program, A2-net [41], uses a CNN framework to identify residues and their poses followed by a Monte Carlo Tree Search (MCTS) strategy to link the 3D coordinate system of amino acids into a complete peptide chain. Compared to its automated counterparts, including Rosetta and others [42][43], A2-net converges to a solution significantly faster. An atomic model for a protein with one thousand amino acids can be derived in minutes [40], compared to several hundred hours in Rosetta [43]. The latest version of A2-net, CryoNet (https://cryonet.ai (accessed on 10 May 2023)), has been applied for model building of cryo-EM maps of the human minor spliceosome [44] and the Acidobacteria homodimeric reaction center bound with cytochrome c [45]. CR-I-TASSER [38] is a fully automated, hybrid method that combines a 3D-CNN to build Cα trace models with multithreading algorithms to identify homologous Protein Data Bank (PDB) templates for guided structure assembly. Nonetheless, the generated CR-I-TASSER model directly relies on the accuracy of the Cα trace prediction, which authors found to decrease with lower resolution maps [38]. Additionally, this technique requires prior segmentation of the experimental map, further limiting its applicability to low-resolution density maps. Although CNNs remain one of the most popular AI architectures for atomic model building and are the most direct DL method for learning features from density maps, they are not without limitations. While CNNs are considered translation-equivariant, meaning translating the input data will result in translated output, they lack rotational invariance, meaning they do not behave consistently for input data of varying orientation. Moreover, the localization of the convolution mechanism narrows the CNN’s receptive field [46][47], potentially limiting their ability to capture global features and dependencies within cryo-EM maps.
In addition to standard CNN framework, many de novo model building programs have adopted the U-Net architecture (Table 1). The U-Net is a type of CNN commonly used to segment or classify pixels in an input image [48]. In the U-Net architecture, an encoder first downsamples the input image and extracts features, and then a decoder network restores the image back to its original dimensions. Through the downsampling and upsampling operations, U-Nets may be more effective at extracting both low-level and high-level features than standard CNNs. Moreover, U-Nets incorporate so-called skip connections that, as the name implies, may skip any given layer in the neural network to provide direct connections between different layers of the network. In the U-Net framework, skip connections enable the direct propagation of feature maps from the encoder to the decoder. This unique architecture preserves fine-grained details and facilitates better information flow throughout the network. Various programs have employed a 3D U-Net framework for SSE identification [49][50] and complete de novo model building [51][52]. DeepTracer [51] is one of the most popular, fully automated model building programs that utilizes a 3D U-Net architecture to predict locations of SSE elements, backbone atoms, and individual amino acids within EM maps. Evolved from Cascaded-CNN [37], DeepTracer employs a series of four U-Nets to perform distinct tasks: locate amino acid positions, locate the protein backbone, identify SSEs, and identify individual residues [51]. Pfab et al. applied DeepTracer to a set of coronavirus-related cryo-EM maps (Figure 2), and found that, on average, 84% residues match with the corresponding deposited PDB structures [51]. Furthermore, DeepTracer has been utilized to build a variety of atomic models from EM maps, including the human small subunit processome [53], Chikungunya virus replication complex [54], and human caveolin-1 complex [55]. While DeepTracer demonstrates more accurate Cα prediction than its counterparts [56][57][58], it is currently only applicable to maps with 5 Å resolution or higher. The latest version, DeepTracer-2.0 [59], also has the ability to model nucleic acids.
Figure 2. The atomic model built from the cryo-EM map of a feline coronavirus spike protein [60] using DeepTracer [51]. (A) The 3.3 Å cryo-EM map of a feline coronavirus spike protein (EMDB ID: EMD-9891) that contains 1403 residues. (B) Density map fitted with the DeepTracer model. The DeepTracer model was built in just 14 min, compared to over 60 h required for model building with Phenix [51]. (C) Visualization of individual backbone atoms and side chains fitted to the cryo-EM Coulomb map using the molecular model obtained with DeepTracer.
Other DL architectures for de novo model building include the Graph Neural Network (GNN), Recurrent Neural Network (RNN), Residual Neural Network (ResNet), and Long- and Short-Term Memory network (LSTM) (Table 1). For example, Structure Generator [39] is a fully automated pipeline that utilizes 3D-CNN, GCN, and LSTM algorithms to produce a protein structure based on amino acid identities and locations. Structure Generator [39] first uses a 3D-CNN, named RotamerNet, to locate amino acids and assign their rotameric orientations within the map. Using the Cα coordinates defined by RotamerNet, the GCN generates a contact map and connects Cα positions located within 4 Å. Lastly, the bidirectional LSTM labels the candidate amino acids and ensures the consistency of the assignment with a provided protein sequence to yield a structural model [39]. While Structure Generator has demonstrated high prediction accuracy using simulated datasets [39], it has not yet been applied to experimental maps. Additionally, the program is limited to modeling protein sequences with less than 700 amino acids [39].
DL techniques have provided much needed insight into the automation of atomic model building. Several approaches have demonstrated capabilities to construct full atomic models, but their applicability is generally limited to maps with 5 Å or higher resolution [51][59][61][62]. Furthermore, while several methodologies are able to precisely identify and locate individual atoms, it remains a challenge to construct a full peptide chain without violating geometric and stereochemical restraints [32][63]. Different solutions have been proposed to remedy this problem, including the program ModelAngelo [61][62] that utilizes sequence data and prior information of protein geometries to refine the protein chain geometry. Another software, CR-I-TASSER, attempts to overcome this limitation by integrating Molecular Dynamics simulations with DL model building [38]. New strategies have also been proposed to integrate DL frameworks with protein structure prediction techniques, such as AlphaFold2, discussed in detail in the following section.
Table 1. Summary of recently developed AI methods for de novo atomic model building.
Program AI Architecture Advantages Limitations
Emap2Sec [34] CNN Identifies SSEs

Demonstrated on experimental maps

High accuracy for intermediate resolution maps
Does not place α-helices and β-sheets in detected regions

Voxel-based approach fails for large EM maps
Emap2Sec+ [36] ResNet Identifies SSEs and nucleic acids in maps 5–10 Å

Improved accuracy in protein SSE detection compared to Emap2Sec [34]
Voxel-based approach fails for large EM maps
CNN-based [35] CNN Identifies SSEs Does not place α-helices and β-sheets in detected regions

No results for experimental maps
Haruspex [49] U-Net Identifies SSEs and nucleic acids in maps

Applied to experimental and simulated maps
Only applicable to maps ≤4 Å

False positives for helices, sheets or RNA/DNA

Misclassifies semi-helical elements, β-hairpin turns, polyproline residues as α-helices
EMNUSS [50] U-Net Identifies SSEs

Applied to experimental and simulated maps
Incorrect predictions on atypical density volumes; narrow receptive field
AAnchor [40] CNN Identifies amino acids Only applicable to maps ≤3.1 Å

Only detects 10–20% amino acids on average
A2-Net [41] 3D-CNN, MCTS Identifies amino acids

Model with 1000 amino acids can be derived in minutes

Applied to experimental maps

Fully automated
Cannot identify ligands
Structure Generator [39] 3D-CNN,

GCN, LSTM
Identifies amino acids and rotamer orientation, builds full protein chain Limited to protein sequences <700 amino acids

No results for experimental maps
CR-I-TASSER [38] 3D-CNN, I-TASSER 3D-CNN predicts Cα atoms used for I-TASSER; generates full structure Cα trace prediction accuracy dependent on resolution

Requires prior map segmentation
DeepTracer [51][59] 3D U-

Net
Locates amino acid positions and protein backbone

Identifies SSEs and amino acids

Automated

Lastest version models nucleic acids
Only applicable to maps ≤5 Å

Only builds atoms for main chains
DeepTracer ID [64] 3D U-Net, AlphaFold2 Uses DeepTracer to build model and searches against AlphaFold2 library for refinement

Identifies individual proteins in density map; does not require protein sequences to be known a priori

Does not require high accuracy from AlphaFold2 prediction
Limited to proteins >100 amino acids for succesful AlphaFold2 prediction matches

Only applicable to maps ≤4.2 Å
EMBUILD [52] U-Net, AlphaFold2 Constructs main chain map; fits AlphaFold2 predicted chains into the map Only builds atoms for main chains
ModelAngelo [61][62] CNN, GNN Builds complete atomic model

Better RMSD and sequence prediction results than DeepTracer
Only applicable to maps ≤3.5 Å

Requires protein sequences are known
DeepMM [65] CNN Predicts Cα positons; identifies SSEs and amino acids

Applied to experimental maps
Cannot model nucleic acids

Residue matching accuracy dependent on map resolution
DEMO-EM [66] ResNET, I-TASSER Builds complete atomic model for multi-domain proteins

Only requires protein sequence

Fully automated
Requires prior map segmentation

Individual domain models calculated without restraints from density data
The field of protein structure prediction has been transformed by significant developments in DL algorithms, exemplified by AlphaFold2 [67], RoseTTaFold [68], ESMFold [69], and I-TASSER-MTD [70]. Given only protein sequences, such programs employ a combination of NNs and novel algorithms to predict atomic 3D structures, which in turn provide complementary data to experimental methods. Of particular note, AlphaFold2 [67], a publicly available program developed by Google DeepMind, has shown unprecedented levels of accuracy in predicting atomic models [67][71][72]. AlphaFold2 outperformed all other prediction methods in the 14th edition of the Critical Assessment of Structure Prediction (CASP), a blind test for structure prediction of experimentally solved structures that have not yet been made publicly available. AlphaFold2 models had a median backbone accuracy of 0.96 Å RMSD, whereas the next best performing method had a median backbone accuracy of 2.80 Å RMSD [67]. Using the amino acid sequence as input, AlphaFold2 first performs a Multiple Sequence Alignment (MSA) and tries to identify proteins with similar structures, known as templates. Based on the templates, AlphaFold2 generates a “pair representation” model, indicating which amino acids are likely to be in contact with each other [73]. The program then employs a transformer NN, called the Evoformer, to refine, exchange, and extract information from the MSA and pair representation. The extracted information is used by the structure module, a second NN, to construct an atomic model [67]. The AlphaFold Protein Structure Database [74], hosted at the European Bioinformatics Institute–European Molecular Biology Laboratory (EMBL-EBI), provides access to over 200 million AlphaFold2 models, covering the majority of sequence entries in UniProt. Furthermore, computed models using AlphaFold2 as well as other prediction methods are available through ModelArchive (modelarchive.org). As of 2022, over one million computed models were also accessible through the Research Collaboratory for Structural Bioinformatics Protein Data Bank (RCSB PDB) [75].
Predicted structures can serve as starting models for deriving atomic coordinates from cryo-EM maps [76][77][78][79][80]. In addition, several methods have been proposed to dock and fit predicted structures into density maps [52][81][82]. In cases of multi-domain and multi-component complexes, one approach is to first use AI-prediction tools to generate structures of sub-complexes and then assemble them into higher-order structures [70][83][84]. For example, the protein structure prediction and modeling tool I-TASSER-MTD [70] predicts structures of individual domains and then uses the EM density to guide the construction of a complete atomic model. Assembline [84] is another software package that offers an integrative pipeline for model building that satisfies the constraints of predicted structures and cryo-EM maps. Assembline has been used to assemble full atomic structures of multiple protein complexes [80][85][86], including the human nuclear core complex [80]. In this restudyarch, Mosalaganti et al. fit models of individual subunits and sub-complexes generated by AlphaFold2 [67] and ColabFold [87], respectively, into the density map, followed by structure assembly with Assembline [84]. Notably, the model of the N-terminal domain of NUP358 predicted with AlphaFold2 was in better agreement with the cryo-EM map than atomic coordinates from previously determined X-ray structures [80]. Furthermore, the predicted models not only showed the same subunit interactions previously reported by the crystal structures, but also revealed new interactions [80].
Another approach is to combine structure prediction tools with DL-based model building techniques to aid in the construction and refinement of atomic models [52][61][63][64][66][81][82]. The program EMBUILD [52] utilizes a U-Net to first generate a so-called main-chain probability map including potential solutions and probabilities reflecting the agreement of the main chain assignments with experimental densities. AlphaFold2 [67] predicted models of individual chains are then fit into the map, and a scoring function is used to measure how well the fitted chain matches the main-chain probability map. The final model is the combination of fitted chains that yield the highest scoring function among different combinations of fitted positions [52]. He et al. [52] applied EMBUILD to 47 intermediate-resolution maps and built models with an average template modeling score (TM-score) of 0.909 and RMSD of 2.85 Å, outperforming other state-of-the-art methods, including phenix.dock_in_map [88] and gmfit [89][90]. Another program, DEMO-EM [66], utilizes DL-based inter-domain distance maps to facilitate the fully automated assembly and refinement of individual domains into full-length structures. When applied to a benchmark set of multi-domain proteins with medium- to low-resolution maps, DEMO-EM generated models with correct inter-domain orientations in 97% of cases, outperforming MDFF and Rosetta [66]. Nonetheless, DEMO-EM, along with the other programs described in this section, utilize models generated for each subunit independently without taking into consideration the conformational changes that may occur upon subunit binding [61].
Other algorithms leverage structure prediction tools to identify proteins within cryo-EM maps. With advancements in image processing techniques, cryo-EM can now be applied directly to cellular extracts. However, determining the specific proteins within the resulting map often requires the application of other complementary techniques, such as tandem Mass Spectrometry. The program DeepTracer-ID [64] combines DeepTracer [51][59] and AlphaFold2 to directly identify the proteins present in a cryo-EM map. DeepTracer-ID first utilizes DeepTracer, as described above, to build an atomic structure, and then uses this model to search the AlphaFold2 library containing all predicted structures from any given organism. Identified structures are used to iteratively refine the atomic model. In a blind test of 13 experimental maps, DeepTracer-ID identified the correct proteins as top candidates [64]. DeepTracer-ID has also been successfully used for protein identification in cryo-EM maps of a white spot syndrome virus capsid [91] and archaeal surface filaments [92]. Nonetheless, this approach is only applicable for maps with 4.2 Å or higher resolution. In addition, accurate sequence-based alignment of the DeepTracer model and AlphaFold2 predicted structure requires proteins with a sequence longer than 100 amino acids [64]. Similarly, the program DeepProLigand [63] predicts protein ligand interactions by employing DeepTracer and using other known protein-ligand structures available in AlphaFold2 library or available through the RCSB PDB.
While integrative approaches described in this section represent a promising tool for atomic model building, AI structure prediction methodologies suffer from several limitations. Assessing the quality of predicted AlphaFold2 structures is a nontrivial task. Although quality metrics accompany each AlphaFold2 model, these scores are also predictions that may contain errors [93]. Thus, manual inspection is still required to evaluate the model in the context of the experimental map, as even high-confidence predictions can be modeled incorrectly [94]. Furthermore, as AlphaFold2 relies on experimentally determined structures deposited in the PDB, the algorithm may produce incorrect folds when there is a lack of homological structures [95]. Further developments are needed to improve the prediction accuracy of mutated residues, regions involved in ligand binding, and dynamic interactions [93][94].

References

  1. DiIorio, M.C.; Kulczyk, A.W. Exploring the Structural Variability of Dynamic Biological Complexes by Single-Particle Cryo-Electron Microscopy. Micromachines 2022, 14, 118.
  2. DiIorio, M.C.; Kulczyk, A.W. A Robust Single-Particle Cryo-Electron Microscopy (cryo-EM) Processing Workflow with cryoSPARC, RELION, and Scipion. J. Vis. Exp. 2022, 179, e63387.
  3. Kulczyk, A.W.; Moeller, A.; Meyer, P.; Sliz, P.; Richardson, C.C. Cryo-EM structure of the replisome reveals multiple interactions coordinating DNA synthesis. Proc. Natl. Acad. Sci. USA 2017, 114, E1848–E1856.
  4. Kulczyk, A.W.; Sorzano, C.O.S.; Grela, P.; Tchorzewski, M.; Tumer, N.E.; Li, X.P. Cryo-EM structure of Shiga toxin 2 in complex with the native ribosomal P-stalk reveals residues involved in the binding interaction. J. Biol. Chem. 2023, 299, 102795.
  5. Burley, S.K.; Berman, H.M.; Chiu, W.; Dai, W.; Flatt, J.W.; Hudson, B.P.; Kaelber, J.T.; Khare, S.D.; Kulczyk, A.W.; Lawson, C.L.; et al. Electron microscopy holdings of the Protein Data Bank: The impact of the resolution revolution, new validation tools, and implications for the future. Biophys. Rev. 2022, 14, 1281–1301.
  6. Kulczyk, A.W.; McKee, K.K.; Zhang, X.M.; Bizukojc, I.; Yu, Y.Q.; Yurchenco, P.D. Cryo-EM reveals the molecular basis of laminin polymerization and LN-lamininopathies. Nat. Commun. 2023, 14, 317.
  7. Kuhlbrandt, W. The resolution revolution. Science 2014, 343, 1443–1444.
  8. Earl, L.A.; Falconieri, V.; Milne, J.L.; Subramaniam, S. Cryo-EM: Beyond the microscope. Curr. Opin. Struct. Biol. 2017, 46, 71–78.
  9. Bai, X.C.; McMullan, G.; Scheres, S.H.W. How cryo-EM is revolutionizing structural biology. Trends Biochem. Sci. 2015, 40, 49–57.
  10. Zhao, J.; Benlekbir, S.; Rubinstein, J.L. Electron cryomicroscopy observation of rotational states in a eukaryotic V-ATPase. Nature 2015, 521, 241–245.
  11. Fica, S.M.; Nagai, K. Cryo-electron microscopy snapshots of the spliceosome: Structural insights into a dynamic ribonucleoprotein machine. Nat. Struct. Mol. Biol. 2017, 24, 791–799.
  12. Wong, W.; Bai, X.C.; Brown, A.; Fernandez, I.S.; Hanssen, E.; Condron, M.; Tan, Y.H.; Baum, J.; Scheres, S.H. Cryo-EM structure of the Plasmodium falciparum 80S ribosome bound to the anti-protozoan drug emetine. eLife 2014, 3, e03080.
  13. Punjani, A.; Rubinstein, J.L.; Fleet, D.J.; Brubaker, M.A. cryoSPARC: Algorithms for rapid unsupervised cryo-EM structure determination. Nat. Methods 2017, 14, 290–296.
  14. Scheres, S.H. RELION: Implementation of a Bayesian approach to cryo-EM structure determination. J. Struct. Biol. 2012, 180, 519–530.
  15. de la Rosa-Trevin, J.M.; Quintana, A.; Del Cano, L.; Zaldivar, A.; Foche, I.; Gutierrez, J.; Gomez-Blanco, J.; Burguet-Castell, J.; Cuenca-Alba, J.; Abrishami, V.; et al. Scipion: A software framework toward integration, reproducibility and validation in 3D electron microscopy. J. Struct. Biol. 2016, 195, 93–99.
  16. Grant, T.; Rohou, A.; Grigorieff, N. cisTEM, user-friendly software for single-particle image processing. eLife 2018, 7, e35383.
  17. Shaikh, T.R.; Gao, H.; Baxter, W.T.; Asturias, F.J.; Boisset, N.; Leith, A.; Frank, J. SPIDER image processing for single-particle reconstruction of biological macromolecules from electron micrographs. Nat. Protoc. 2008, 3, 1941–1974.
  18. van Heel, M.; Harauz, G.; Orlova, E.V.; Schmidt, R.; Schatz, M. A new generation of the IMAGIC image processing system. J. Struct. Biol. 1996, 116, 17–24.
  19. Ludtke, S.J.; Baldwin, P.R.; Chiu, W. EMAN: Semiautomated software for high-resolution single-particle reconstructions. J. Struct. Biol. 1999, 128, 82–97.
  20. Lander, G.C.; Stagg, S.M.; Voss, N.R.; Cheng, A.; Fellmann, D.; Pulokas, J.; Yoshioka, C.; Irving, C.; Mulder, A.; Lau, P.W.; et al. Appion: An integrated, database-driven pipeline to facilitate EM image processing. J. Struct. Biol. 2009, 166, 95–102.
  21. Baldwin, P.R.; Tan, Y.Z.; Eng, E.T.; Rice, W.J.; Noble, A.J.; Negro, C.J.; Cianfrocco, M.A.; Potter, C.S.; Carragher, B. Big data in cryoEM: Automated collection, processing and accessibility of EM data. Curr. Opin. Microbiol. 2018, 43, 1–8.
  22. Namba, K.; Makino, F. Recent progress and future perspective of electron cryomicroscopy for structural life sciences. Microscopy 2022, 71, i3–i14.
  23. Penczek, P.A.; Fang, J.; Li, X.; Cheng, Y.; Loerke, J.; Spahn, C.M. CTER-rapid estimation of CTF parameters with error assessment. Ultramicroscopy 2014, 140, 9–19.
  24. Heimowitz, A.; Anden, J.; Singer, A. Reducing bias and variance for CTF estimation in single particle cryo-EM. Ultramicroscopy 2020, 212, 112950.
  25. Arbelaez, P.; Han, B.G.; Typke, D.; Lim, J.; Glaeser, R.M.; Malik, J. Experimental evaluation of support vector machine-based and correlation-based approaches to automatic particle selection. J. Struct. Biol. 2011, 175, 319–328.
  26. Heimowitz, A.; Anden, J.; Singer, A. APPLE picker: Automatic particle picking, a low-effort cryo-EM framework. J. Struct. Biol. 2018, 204, 215–227.
  27. Al-Azzawi, A.; Ouadou, A.; Tanner, J.J.; Cheng, J. AutoCryoPicker: An unsupervised learning approach for fully automated single particle picking in Cryo-EM images. BMC Bioinform. 2019, 20, 326.
  28. Yang, Z.; Fang, J.; Chittuluru, J.; Asturias, F.J.; Penczek, P.A. Iterative stable alignment and clustering of 2D transmission electron microscope images. Structure 2012, 20, 237–247.
  29. Ma, L.; Reisert, M.; Burkhardt, H. RENNSH: A novel alpha-helix identification approach for intermediate resolution electron density maps. IEEE/ACM Trans. Comput. Biol. Bioinform. 2012, 9, 228–239.
  30. Chen, M.; Baldwin, P.R.; Ludtke, S.J.; Baker, M.L. De Novo modeling in cryo-EM density maps with Pathwalking. J. Struct. Biol. 2016, 196, 289–298.
  31. Si, D.; Ji, S.; Nasr, K.A.; He, J. A machine learning approach for the identification of protein secondary structure elements from electron cryo-microscopy density maps. Biopolymers 2012, 97, 698–708.
  32. Giri, N.; Roy, R.S.; Cheng, J. Deep learning for reconstructing protein structures from cryo-EM density maps: Recent advances and future directions. Curr. Opin. Struct. Biol. 2023, 79, 102536.
  33. Si, D.; Nakamura, A.; Tang, R.B.; Guan, H.W.; Hou, J.; Firozi, A.; Cao, R.Z.; Hippe, K.; Zhao, M.L. Artificial intelligence advances for de novo molecular structure modeling in cryo-electron microscopy. WIREs Comput. Mol. Sci. 2022, 12, e1542.
  34. Subramaniya, S.R.M.V.; Terashi, G.; Kihara, D. Protein Secondary Structure Detection in Intermediate-Resolution Cryo-EM Maps using Deep Learning. Nat. Methods 2020, 118, 43a.
  35. Li, R.J.; Si, D.; Zeng, T.; Ji, S.W.; He, J. Deep Convolutional Neural Networks for Detecting Secondary Structures in Protein Density Maps from Cryo-Electron Microscopy. In Proceedings of the IEEE International Conference on Bioinformatics and Biomedicine (BIBM), Shenzhen, China, 15–18 December 2016; pp. 41–46.
  36. Wang, X.; Alnabati, E.; Aderinwale, T.W.; Subramaniya, S.R.M.V.; Terashi, G.; Kihara, D. Emap2sec+: Detecting Protein and DNA/RNA Structures in Cryo-EM Maps of Intermediate Resolution Using Deep Learning. Acta Crystallogr. Sect. A 2021, 77, A84.
  37. Si, D.; Moritz, S.A.; Pfab, J.; Hou, J.; Cao, R.Z.; Wang, L.G.; Wu, T.Q.; Cheng, J.L. Deep Learning to Predict Protein Backbone Structure from High-Resolution Cryo-EM Density Maps. Sci. Rep. 2020, 10, 4282.
  38. Zhang, X.; Zhang, B.; Freddolino, P.L.; Zhang, Y. CR-I-TASSER: Assemble protein structures from cryo-EM density maps using deep convolutional neural networks. Nat. Methods 2022, 19, 195–204.
  39. Li, P.N.; de Oliveira, S.H.P.; Wakatsuki, S.; van den Bedem, H. Sequence-guided protein structure determination using graph convolutional and recurrent networks. In Proceedings of the IEEE 20th International Conference on Bioinformatics and Bioengineering (BIBE), Cincinnati, OH, USA, 26–28 October 2020; pp. 122–127.
  40. Rozanov, M.; Wolfson, H.J. AAnchor: CNN guided detection of anchor amino acids in high resolution cryo-EM density maps. In Proceedings of the IEEE 20th International Conference on Bioinformatics and Bioengineering (BIBE), Madrid, Spain, 3–6 December 2018; pp. 88–91.
  41. Xu, K.; Wang, Z.; Shi, J.P.; Li, H.S.; Zhang, Q.C. A(2)-Net: Molecular Structure Estimation from Cryo-EM Density Volumes. In Proceedings of the AAAI Conference on Artificial Intelligence, Honolulu, HI, USA, 27 January–1 February 2019; pp. 1230–1237.
  42. He, K.M.; Zhang, X.Y.; Ren, S.Q.; Sun, J. Deep Residual Learning for Image Recognition. In Proceedings of the 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Las Vegas, NV, USA, 26 June–1 July 2016; pp. 770–778.
  43. Wang, R.Y.R.; Kudryashev, M.; Li, X.M.; Egelman, E.H.; Basler, M.; Cheng, Y.F.; Baker, D.; DiMaio, F. De novo protein structure determination from near-atomic-resolution cryo-EM maps. Nat. Methods 2015, 12, 335–384.
  44. Bai, R.; Wan, R.; Wang, L.; Xu, K.; Zhang, Q.; Lei, J.; Shi, Y. Structure of the activated human minor spliceosome. Science 2021, 371, eabg0879.
  45. Dong, S.; Huang, G.; Wang, C.; Wang, J.; Sui, S.F.; Qin, X. Structure of the Acidobacteria homodimeric reaction center bound with cytochrome c. Nat. Commun. 2022, 13, 7745.
  46. Liu, W.; Rabinovich, A.; Berg, A.C. ParseNet: Looking wider to see better. arXiv 2015, arXiv:1506.04579.
  47. Peng, C.; Zhang, X.Y.; Yu, G.; Luo, G.M.; Sun, J. Large Kernel Matters—Improve Semantic Segmentation by Global Convolutional Network. In Proceedings of the 30th IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Honolulu, HI, USA, 21–26 July 2017; pp. 1743–1751.
  48. Ronneberger, O.; Fischer, P.; Brox, T. U-Net: Convolutional Networks for Biomedical Image Segmentation. In Proceedings of the International Conference on Medical Image Computing and Computer-Assisted Intervention, Munich, Germany, 5–9 October 2015; pp. 234–241.
  49. Mostosi, P.; Schindelin, H.; Kollmannsberger, P.; Thorn, A. Haruspex: A Neural Network for the Automatic Identification of Oligonucleotides and Protein Secondary Structure in Cryo-Electron Microscopy Maps. Angew. Chem. Int. Ed. 2020, 59, 14788–14795.
  50. He, J.H.; Huang, S.Y. EMNUSS: A deep learning framework for secondary structure annotation in cryo-EM maps. Brief. Bioinform. 2021, 22, bbab156.
  51. Pfab, J.; Phan, N.M.; Si, D. DeepTracer for fast de novo cryo-EM protein structure modeling and special studies on CoV-related complexes. Proc. Natl. Acad. Sci. USA 2021, 118, e2017525118.
  52. He, J.; Lin, P.; Chen, J.; Cao, H.; Huang, S.Y. Model building of protein complexes from intermediate-resolution cryo-EM maps with deep learning-guided automatic assembly. Nat. Commun. 2022, 13, 4066.
  53. Singh, S.; Vanden Broeck, A.; Miller, L.; Chaker-Margot, M.; Klinge, S. Nucleolar maturation of the human small subunit processome. Science 2021, 373, eabj5338.
  54. Tan, Y.B.; Chmielewski, D.; Law, M.C.Y.; Zhang, K.; He, Y.; Chen, M.Y.; Jin, J.; Luo, D.H. Molecular architecture of the Chikungunya virus replication complex. Sci. Adv. 2022, 8, eadd2536.
  55. Porta, J.C.; Han, B.; Gulsevin, A.; Chung, J.M.; Peskova, Y.; Connolly, S.; Mchaourab, H.S.; Meiler, J.; Karakas, E.; Kenworthy, A.K.; et al. Molecular architecture of the human caveolin-1 complex. Sci. Adv. 2022, 8, eabn7232.
  56. Terashi, G.; Kihara, D. De novo main-chain modeling for EM maps using MAINMAST. Nat. Commun. 2018, 9, 1618.
  57. Adams, P.D.; Afonine, P.V.; Bunkoczi, G.; Chen, V.B.; Davis, I.W.; Echols, N.; Headd, J.J.; Hung, L.W.; Kapral, G.J.; Grosse-Kunstleve, R.W.; et al. PHENIX: A comprehensive Python-based system for macromolecular structure solution. Acta Crystallogr. Sect. D Biol. Crystallogr. 2010, 66, 213–221.
  58. Frenz, B.; Walls, A.C.; Egelman, E.H.; Veesler, D.; DiMaio, F. RosettaES: A sampling strategy enabling automated interpretation of difficult cryo-EM maps. Nat. Methods 2017, 14, 797–800.
  59. Nakamura, A.; Meng, H.; Zhao, M.; Wang, F.; Hou, J.; Cao, R.; Si, D. Fast and automated protein-DNA/RNA macromolecular complex modeling from cryo-EM maps. Brief. Bioinform. 2023, 24, bbac632.
  60. Yang, T.J.; Chang, Y.C.; Ko, T.P.; Draczkowski, P.; Chien, Y.C.; Chang, Y.C.; Wu, K.P.; Khoo, K.H.; Chang, H.W.; Hsu, S.D. Cryo-EM analysis of a feline coronavirus spike protein reveals a unique structure and camouflaging glycans. Proc. Natl. Acad. Sci. USA 2020, 117, 1438–1446.
  61. Jamali, K.; Kimanius, D.; Scheres, S.H. A Graph Neural Network Approach to Automated Model Building in Cryo-EM Maps. In Proceedings of the International Conference on Learning Representations, Kigali, Rowanda, 1–5 May 2023.
  62. Jamali, K.; Kall, L.; Zhang, R.; Brown, A.; Kimanius, D.; Scheres, S.H. Automated model building and protein identification in cryo-EM maps. bioRxiv 2023.
  63. Giri, N.; Cheng, J.L. Improving Protein-Ligand Interaction Modeling with cryo-EM Data, Templates, and Deep Learning in 2021 Ligand Model Challenge. Biomolecules 2023, 13, 132.
  64. Chang, L.C.; Wang, F.B.; Connolly, K.; Meng, H.Z.; Su, Z.L.; Cvirkaite-Krupovic, V.; Krupovic, M.; Egelman, E.H.; Si, D. DeepTracer-ID: De novo protein identification from cryo-EM maps. Biophys. J. 2022, 121, 2840–2848.
  65. He, J.; Huang, S.Y. Full-length de novo protein structure determination from cryo-EM maps using deep learning. Bioinformatics 2021, 37, 3480–3490.
  66. Zhou, X.; Li, Y.; Zhang, C.; Zheng, W.; Zhang, G.; Zhang, Y. Progressive assembly of multi-domain protein structures from cryo-EM density maps. Nat. Comput. Sci. 2022, 2, 265–275.
  67. Jumper, J.; Evans, R.; Pritzel, A.; Green, T.; Figurnov, M.; Ronneberger, O.; Tunyasuvunakool, K.; Bates, R.; Zidek, A.; Potapenko, A.; et al. Highly accurate protein structure prediction with AlphaFold. Nature 2021, 596, 583–589.
  68. Baek, M.; DiMaio, F.; Anishchenko, I.; Dauparas, J.; Ovchinnikov, S.; Lee, G.R.; Wang, J.; Cong, Q.; Kinch, L.N.; Schaeffer, R.D.; et al. Accurate prediction of protein structures and interactions using a three-track neural network. Science 2021, 373, 871–876.
  69. Lin, Z.; Halil, A.; Rao, R.; Hie, B.; Zhu, Z.; Lu, W.; dos Santos Costa, A.; Fazel-Zarandi, M.; Sercu, T.; Candido, S.; et al. Language models of protein sequences at the scale of evolution enable accurate structure prediction. bioRxiv 2022.
  70. Zhou, X.; Zheng, W.; Li, Y.; Pearce, R.; Zhang, C.; Bell, E.W.; Zhang, G.; Zhang, Y. I-TASSER-MTD: A deep-learning-based platform for multi-domain protein structure and function prediction. Nat. Protoc. 2022, 17, 2326–2353.
  71. Bryant, P.; Pozzati, G.; Elofsson, A. Improved prediction of protein-protein interactions using AlphaFold2. Nat. Commun. 2022, 13, 1265.
  72. Jumper, J.; Evans, R.; Pritzel, A.; Green, T.; Figurnov, M.; Ronneberger, O.; Tunyasuvunakool, K.; Bates, R.; Zidek, A.; Potapenko, A.; et al. Applying and improving AlphaFold at CASP14. Proteins 2021, 89, 1711–1721.
  73. AF2 Is Here: What’s Behind the Structure Prediction Miracle. Available online: https://www.blopig.com/blog/2021/07/alphafold-2-is-here-whats-behind-the-structure-prediction-miracle (accessed on 19 June 2023).
  74. Varadi, M.; Anyango, S.; Deshpande, M.; Nair, S.; Natassia, C.; Yordanova, G.; Yuan, D.; Stroe, O.; Wood, G.; Laydon, A.; et al. AlphaFold Protein Structure Database: Massively expanding the structural coverage of protein-sequence space with high-accuracy models. Nucleic Acids Res. 2022, 50, D439–D444.
  75. Burley, S.K.; Bhikadiya, C.; Bi, C.X.; Bittrich, S.; Chao, H.Y.; Chen, L.; Craig, P.A.; Crichlow, G.V.; Dalenberg, K.; Duarte, J.M.; et al. RCSB Protein Data Bank (RCSB.org): Delivery of experimentally-determined PDB structures alongside one million computed structure models of proteins from artificial intelligence/machine learning. Nucleic Acids Res. 2022, 51, D488–D508.
  76. Andreeva, L.; David, L.; Rawson, S.; Shen, C.; Pasricha, T.; Pelegrin, P.; Wu, H. NLRP3 cages revealed by full-length mouse NLRP3 structure control pathway activation. Cell 2021, 184, 6299–6312.
  77. Noddings, C.M.; Wang, R.Y.; Johnson, J.L.; Agard, D.A. Structure of Hsp90-p23-GR reveals the Hsp90 client-remodelling mechanism. Nature 2022, 601, 465–469.
  78. Jones, M.L.; Baris, Y.; Taylor, M.R.G.; Yeeles, J.T.P. Structure of a human replisome shows the organisation and interactions of a DNA replication machine. EMBO J. 2021, 40, e108819.
  79. Fontana, P.; Dong, Y.; Pi, X.; Tong, A.B.; Hecksel, C.W.; Wang, L.F.; Fu, T.M.; Bustamante, C.; Wu, H. Structure of cytoplasmic ring of nuclear pore complex by integrative cryo-EM and AlphaFold. Science 2022, 376, eabm9326.
  80. Mosalaganti, S.; Obarska-Kosinska, A.; Siggel, M.; Taniguchi, R.; Turonova, B.; Zimmerli, C.E.; Buczak, K.; Schmidt, F.H.; Margiotta, E.; Mackmull, M.T.; et al. AI-based structure prediction empowers integrative structural analysis of human nuclear pores. Science 2022, 376, eabm9506.
  81. Terwilliger, T.C.; Poon, B.K.; Afonine, P.V.; Schlicksup, C.J.; Croll, T.I.; Millan, C.; Richardson, J.S.; Read, R.J.; Adams, P.D. Improved AlphaFold modeling with implicit experimental information. Nat. Methods 2022, 19, 1376–1382.
  82. Alshammari, M.; He, J.; Wriggers, W. Refinement of AlphaFold2 Models against Experimental Cryo-EM Density Maps at 4–6Å Resolution. In Proceedings of the 2022 IEEE International Conference on Bioinformatics and Biomedicine, Las Vegas, NV, USA, 6–8 December 2022; pp. 3423–3430.
  83. Ziemianowicz, D.S.; Kosinski, J. New opportunities in integrative structural modeling. Curr. Opin. Struc. Biol. 2022, 77, 102488.
  84. Rantos, V.; Karius, K.; Kosinski, J. Integrative structural modeling of macromolecular complexes using Assembline. Nat. Protoc. 2022, 17, 152–176.
  85. Beckham, K.S.H.; Ritter, C.; Chojnowski, G.; Ziemianowicz, D.S.; Mullapudi, E.; Rettel, M.; Savitski, M.M.; Mortensen, S.A.; Kosinski, J.; Wilmanns, M. Structure of the mycobacterial ESX-5 type VII secretion system pore complex. Sci. Adv. 2021, 7, eabg9923.
  86. Flacht, L.; Lunelli, M.; Kaszuba, K.; Chen, Z.A.; Reilly, F.J.O.; Rappsilber, J.; Kosinski, J.; Kolbe, M. Integrative structural analysis of the type III secretion system needle complex from Shigella flexneri. Protein Sci. 2023, 32, e4595.
  87. Mirdita, M.; Schutze, K.; Moriwaki, Y.; Heo, L.; Ovchinnikov, S.; Steinegger, M. ColabFold: Making protein folding accessible to all. Nat. Methods 2022, 19, 679–682.
  88. Liebschner, D.; Afonine, P.V.; Baker, M.L.; Bunkóczi, G.; Chen, V.B.; Croll, T.I.; Hintze, B.; Hung, L.W.; Jain, S.; McCoy, A.J.; et al. Macromolecular structure determination using X-rays, neutrons and electrons: Recent developments in Phenix. Acta Crystallogr. Sect. D Struct. Biol. 2019, 75, 861–877.
  89. Kawabata, T. Multiple subunit fitting into a low-resolution density map of a macromolecular complex using a gaussian mixture model. Biophys. J. 2008, 95, 4643–4658.
  90. Kawabata, T. Gaussian-input Gaussian mixture model for representing density maps and atomic models. J. Struct. Biol. 2018, 203, 1–16.
  91. Sun, M.L.; Liu, M.D.; Shan, H.; Li, K.; Wang, P.; Guo, H.R.; Zhao, Y.Q.; Wang, R.; Tao, Y.W.; Yang, L.Y.; et al. Ring-stacked capsids of white spot syndrome virus and structural transitions with genome ejection. Sci. Adv. 2023, 9, eadd2796.
  92. Wang, F.B.; Cvirkaite-Krupovic, V.; Krupovic, M.; Egelman, E.H. Archaeal bundling pili of Pyrobaculum calidifontis reveal similarities between archaeal and bacterial biofilms. Proc. Natl. Acad. Sci. USA 2022, 119, e2216660119.
  93. Allison, T.M.; Degiacomi, M.T.; Marklund, E.G.; Jovine, L.; Elofsson, A.; Benesch, J.L.P.; Landreh, M. Complementing machine learning-based structure predictions with native mass spectrometry. Protein Sci. 2022, 31, e4333.
  94. Akdel, M.; Pires, D.E.V.; Pardo, E.P.; Janes, J.; Zalevsky, A.O.; Meszaros, B.; Bryant, P.; Good, L.L.; Laskowski, R.A.; Pozzati, G.; et al. A structural biology community assessment of AlphaFold2 applications. Nat. Struct. Mol. Biol. 2022, 29, 1056–1067.
  95. Si, D.; Chen, J.; Nakamura, A.; Chang, L.; Guan, H. Smart de novo Macromolecular Structure Modeling from Cryo-EM Maps. J. Mol. Biol. 2023, 435, 167967.
More
ScholarVision Creations