Cryo-EM model validation recommendations based on outcomes of the 2019 EMDataResource challenge

Atomic model validation using the CCP-EM software suite

Recently, there has been a dramatic improvement in the quality and quantity of data derived using cryogenic electron microscopy (cryo-EM). This is also associated with a large increase in the number of atomic models built. Although the best resolutions that are achievable are improving, often the local resolution is variable, and a significant majority of data are still resolved at resolutions worse than 3 Å. Model building and refinement is often challenging at these resolutions, and hence atomic model validation becomes even more crucial to identify less reliable regions of the model. Here, a graphical user interface for atomic model validation, implemented in the CCP-EM software suite, is presented. It is aimed to develop this into a platform where users can access multiple complementary validation metrics that work across a range of resolutions and obtain a summary of evaluations. Based on the validation estimates from atomic models associated with cryo-EM structures from SARS-CoV-2, it was observed that models typically favor adopting the most common conformations over fitting the observations when compared with the model agreement with data. At low resolutions, the stereochemical quality may be favored over data fit, but care should be taken to ensure that the model agrees with the data in terms of resolvable features. It is demonstrated that further re-refinement can lead to improvement of the agreement with data without the loss of geometric quality. This also highlights the need for improved resolution-dependent weight optimization in model refinement and an effective test for overfitting that would help to guide the refinement process.

findMySequence: a neural-network-based approach for identification of unknown proteins in X-ray crystallography and cryo-EM

Although experimental protein-structure determination usually targets known proteins, chains of unknown sequence are often encountered. They can be purified from natural sources, appear as an unexpected fragment of a well characterized protein or appear as a contaminant. Regardless of the source of the problem, the unknown protein always requires characterization. Here, an automated pipeline is presented for the identification of protein sequences from cryo-EM reconstructions and crystallographic data. The method's application to characterize the crystal structure of an unknown protein purified from a snake venom is presented. It is also shown that the approach can be successfully applied to the identification of protein sequences and validation of sequence assignments in cryo-EM protein structures.

Cryo-EM targets in CASP14

Structures of seven CASP14 targets were determined using cryo-electron microscopy (cryo-EM) technique with resolution between 2.1 and 3.8 Å. We provide an evaluation of the submitted models versus the experimental data (cryo-EM density maps) and experimental reference structures built into the maps. The accuracy of models is measured in terms of coordinate-to-density and coordinate-to-coordinate fit. A-posteriori refinement of the most accurate models in their corresponding cryo-EM density resulted in structures that are close to the reference structure, including some regions with better fit to the density. Regions that were found to be less "refineable" correlate well with regions of high diversity between the CASP models and low goodness-of-fit to density in the reference structure.

High-Performance Deep Learning Toolbox for Genome-Scale Prediction of Protein Structure and Function

Computational biology is one of many scientific disciplines ripe for innovation and acceleration with the advent of high-performance computing (HPC). In recent years, the field of machine learning has also seen significant benefits from adopting HPC practices. In this work, we present a novel HPC pipeline that incorporates various machine-learning approaches for structure-based functional annotation of proteins on the scale of whole genomes. Our pipeline makes extensive use of deep learning and provides computational insights into best practices for training advanced deep-learning models for high-throughput data such as proteomics data. We showcase methodologies our pipeline currently supports and detail future tasks for our pipeline to envelop, including large-scale sequence comparison using SAdLSA and prediction of protein tertiary structures using AlphaFold2.

Electron Diffraction of Ionic Argon Nanoclusters Embedded in Superfluid Helium Droplets

We report electron diffraction of cationic argon nanoclusters embedded in superfluid helium droplets. Superfluid helium droplets are first doped with neutral argon atoms to form nanoclusters, and then the doped droplets are ionized by electrons. The much lower ionization energy of argon ensures that the positive charge resides on the Ar nanocluster. Using different stagnation temperatures and therefore droplets with different sizes, we have been able to preferentially form a small ionic cluster containing 2-4 Ar atoms and a larger cluster containing 7-11 atoms. The fitting results of the diffraction profiles agree with structures reported from theoretical calculations, containing a cationic trimer core with the remaining atoms largely neutral. This work testifies to the feasibility of performing electron diffraction from ionic species embedded in superfluid helium droplets, dispelling the concern over the particle density in the diffraction region. However, the large number of neutral helium atoms surrounding the cationic nanoclusters poses a challenge for the detection of the helium solvation layer, and the detection of which awaits further technological improvements.

CryoFold: determining protein structures and data-guided ensembles from cryo-EM density maps

Cryo-electron microscopy (EM) requires molecular modeling to refine structural details from data. Ensemble models arrive at low free-energy molecular structures, but are computationally expensive and limited to resolving only small proteins that cannot be resolved by cryo-EM. Here, we introduce CryoFold - a pipeline of molecular dynamics simulations that determines ensembles of protein structures directly from sequence by integrating density data of varying sparsity at 3-5 Å resolution with coarse-grained topological knowledge of the protein folds. We present six examples showing its broad applicability for folding proteins between 72 to 2000 residues, including large membrane and multi-domain systems, and results from two EMDB competitions. Driven by data from a single state, CryoFold discovers ensembles of common low-energy models together with rare low-probability structures that capture the equilibrium distribution of proteins constrained by the density maps. Many of these conformations, unseen by traditional methods, are experimentally validated and functionally relevant. We arrive at a set of best practices for data-guided protein folding that are controlled using a Python GUI.

Predicted pH-dependent stability of SARS-CoV-2 spike protein trimer from interfacial acidic groups

Transition between receptor binding domain (RBD) up and down forms of the SARS-CoV-2 spike protein trimer is coupled to receptor binding and is one route by which variants can alter viral properties. It is becoming apparent that key roles in the transition are played by pH and a more compact closed form, termed locked. Calculations of pH-dependence are made for a large set of spike trimers, including locked form trimer structures that have recently become available. Several acidic sidechains become sufficiently buried in the locked form to give a predicted pH-dependence in the mild acidic range, with stabilisation of the locked form as pH reduces from 7.5 to 5, consistent with emerging characterisation by cryo-electron microscopy. The calculated pH effects in pre-fusion spike trimers are modulated mainly by aspartic acid residues, rather than the more familiar histidine role at mild acidic pH. These acidic sidechains are generally surface located and weakly interacting when not in a locked conformation. According to this model, their replacement (perhaps with asparagine) would remove the pH-dependent destabilisation of locked spike trimer conformations, and increase their recovery at neutral pH. This would provide an alternative or supplement to the insertion of disulphide linkages for stabilising spike protein trimers, with potential relevance for vaccine design.

Validation, analysis and annotation of cryo-EM structures

The process of turning 2D micrographs into 3D atomic models of the imaged macromolecules has been under rapid development and scrutiny in the field of cryo-EM. Here, some important methods for validation at several stages in this process are described. Firstly, how Fourier shell correlation of two independent maps and phase randomization beyond a certain frequency address the assessment of map resolution is reviewed. Techniques for local resolution estimation and map sharpening are also touched upon. The topic of validating models which are either built de novo or based on a known atomic structure fitted into a cryo-EM map is then approached. Map-model comparison using Q-scores and Fourier shell correlation plots is used to assure the agreement of the model with the observed map density. The importance of annotating the model with B factors to account for the resolvability of individual atoms in the map is illustrated. Finally, the timely topic of detecting and validating water molecules and metal ions in maps that have surpassed ∼2 Å resolution is described.

Cryo-EM Map-Based Model Validation Using the False Discovery Rate Approach

Significant technological developments and increasing scientific interest in cryogenic electron microscopy (cryo-EM) has resulted in a rapid increase in the amount of data generated by these experiments and the derived atomic models. Robust measures for the validation of 3D reconstructions and atomic models are essential for appropriate interpretation of the data. The resolution of data and availability of software tools that work across a range of resolutions often limit the quality of derived models. Hence, the final atomic model is often incomplete or contains regions where atomic positions are less reliable or incorrectly built. Extensive manual pruning and local adjustments or rebuilding are usually required to address these issues. The presented research introduces a software tool for the validation of the backbone trace of atomic models built in the cryo-EM density maps. In this study, we use the false discovery rate analysis, which can be used to segregate molecular signals from the background. Each atomic position in the model can be associated with an FDR backbone validation score, which can be used to identify potential mistraced residues. We demonstrate that the proposed validation score is complementary to existing validation metrics and is useful especially in cases where the model is built in the maps having varying local resolution. We also discuss the application of the score for automated pruning of atomic models built ab-initio during the iterative model building process in Buccaneer. We have implemented this score in the CCP-EM software suite.

Full-length de novo protein structure determination from cryo-EM maps using deep learning

MOTIVATION

Advances in microscopy instruments and image processing algorithms have led to an increasing number of cryo-EM maps. However, building accurate models for the EM maps at 3-5 Å resolution remains a challenging and time-consuming process. With the rapid growth of deposited EM maps, there is an increasing gap between the maps and reconstructed/modeled 3-dimensional (3D) structures. Therefore, automatic reconstruction of atomic-accuracy full-atomstructures fromEMmaps is pressingly needed.

RESULTS

We present a semi-automatic de novo structure determination method using a deep learningbased framework, named as DeepMM, which builds atomic-accuracy all-atom models from cryo-EM maps at near-atomic resolution. In our method, the main-chain and Cα positions as well as their amino acid and secondary structure types are predicted in the EM map using Densely Connected Convolutional Networks. DeepMM was extensively validated on 40 simulated maps at 5 Å resolution and 30 experimental maps at 2.6-4.8 Å resolution as well as an EMDB-wide data set of 2931 experimental maps at 2.6-4.9 Å resolution, and compared with state-of-the-art algorithms including RosettaES, MAINMAST, and Phenix. Overall, our DeepMM algorithm obtained a significant improvement over existing methods in terms of both accuracy and coverage in building full-length protein structures on all test sets, demonstrating the efficacy and general applicability of DeepMM.

AVAILABILITY

http://huanglab.phys.hust.edu.cn/DeepMM.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

Evolution of standardization and dissemination of cryo-EM structures and data jointly by the community, PDB, and EMDB

Cryogenic electron microscopy (cryo-EM) methods began to be used in the mid-1970s to study thin and periodic arrays of proteins. Following a half-century of development in cryo-specimen preparation, instrumentation, data collection, data processing, and modeling software, cryo-EM has become a routine method for solving structures from large biological assemblies to small biomolecules at near to true atomic resolution. This review explores the critical roles played by the Protein Data Bank (PDB) and Electron Microscopy Data Bank (EMDB) in partnership with the community to develop the necessary infrastructure to archive cryo-EM maps and associated models. Public access to cryo-EM structure data has in turn facilitated better understanding of structure–function relationships and advancement of image processing and modeling tool development. The partnership between the global cryo-EM community and PDB and EMDB leadership has synergistically shaped the standards for metadata, one-stop deposition of maps and models, and validation metrics to assess the quality of cryo-EM structures. The advent of cryo-electron tomography (cryo-ET) for in situ molecular cell structures at a broad resolution range and their correlations with other imaging data introduce new data archival challenges in terms of data size and complexity in the years to come.

Improving SARS-CoV-2 structures: Peer review by early coordinate release

This work builds upon the record-breaking speed and generous immediate release of new experimental three-dimensional structures of the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) proteins and complexes, which are crucial to downstream vaccine and drug development. We have surveyed those structures to catch the occasional errors that could be significant for those important uses and for which we were able to provide demonstrably higher-accuracy corrections. This process relied on new validation and correction methods such as CaBLAM and ISOLDE, which are not yet in routine use. We found such important and correctable problems in seven early SARS-CoV-2 structures. Two of the structures were soon superseded by new higher-resolution data, confirming our proposed changes. For the other five, we emailed the depositors a documented and illustrated report and encouraged them to make the model corrections themselves and use the new option at the worldwide Protein Data Bank for depositors to re-version their coordinates without changing the Protein Data Bank code. This quickly and easily makes the better-accuracy coordinates available to anyone who examines or downloads their structure, even before formal publication. The changes have involved sequence misalignments, incorrect RNA conformations near a bound inhibitor, incorrect metal ligands, and cis-trans or peptide flips that prevent good contact at interaction sites. These improvements have propagated into nearly all related structures done afterward. This process constitutes a new form of highly rigorous peer review, which is actually faster and more strict than standard publication review because it has access to coordinates and maps; journal peer review would also be strengthened by such access.