Outcomes of the EMDataResource cryo-EM Ligand Modeling Challenge.

The EMDataResource Ligand Model Challenge aimed to assess the reliability and reproducibility of modeling ligands bound to protein and protein-nucleic acid complexes in cryogenic electron microscopy (cryo-EM) maps determined at near-atomic (1.9-2.5 Å) resolution. Three published maps were selected as targets: Escherichia coli beta-galactosidase with inhibitor, SARS-CoV-2 virus RNA-dependent RNA polymerase with covalently bound nucleotide analog and SARS-CoV-2 virus ion channel ORF3a with bound lipid. Sixty-one models were submitted from 17 independent research groups, each with supporting workflow details. The quality of submitted ligand models and surrounding atoms were analyzed by visual inspection and quantification of local map quality, model-to-map fit, geometry, energetics and contact scores. A composite rather than a single score was needed to assess macromolecule+ligand model quality. These observations lead us to recommend best practices for assessing cryo-EM structures of liganded macromolecules reported at near-atomic resolution.

Outcomes of the EMDataResource Cryo-EM Ligand Modeling Challenge.

The EMDataResource Ligand Model Challenge aimed to assess the reliability and reproducibility of modeling ligands bound to protein and protein/nucleic-acid complexes in cryogenic electron microscopy (cryo-EM) maps determined at near-atomic (1.9-2.5 Å) resolution. Three published maps were selected as targets: E. coli beta-galactosidase with inhibitor, SARS-CoV-2 RNA-dependent RNA polymerase with covalently bound nucleotide analog, and SARS-CoV-2 ion channel ORF3a with bound lipid. Sixty-one models were submitted from 17 independent research groups, each with supporting workflow details. We found that (1) the quality of submitted ligand models and surrounding atoms varied, as judged by visual inspection and quantification of local map quality, model-to-map fit, geometry, energetics, and contact scores, and (2) a composite rather than a single score was needed to assess macromolecule+ligand model quality. These observations lead us to recommend best practices for assessing cryo-EM structures of liganded macromolecules reported at near-atomic resolution.

Accounting for nonuniformity of bulk-solvent: A mosaic model.

A flat mask-based model is almost universally used in macromolecular crystallography to account for disordered (bulk) solvent. This model assumes any voxel of the crystal unit cell that is not occupied by the atomic model is occupied by the solvent. The properties of this solvent are assumed to be exactly the same across the whole volume of the unit cell. While this is a reasonable approximation in practice, there are a number of scenarios where this model becomes suboptimal. In this work, we enumerate several of these scenarios and describe a new generalized approach to modeling the bulk-solvent which we refer to as mosaic bulk-solvent model. The mosaic bulk-solvent model allows nonuniform features of the solvent in the crystal to be accounted for in a computationally efficient way. It is implemented in the computational crystallography toolbox and the Phenix software.

AlphaFold predictions are valuable hypotheses and accelerate but do not replace experimental structure determination.

Artificial intelligence-based protein structure prediction methods such as AlphaFold have revolutionized structural biology. The accuracies of these predictions vary, however, and they do not take into account ligands, covalent modifications or other environmental factors. Here, we evaluate how well AlphaFold predictions can be expected to describe the structure of a protein by comparing predictions directly with experimental crystallographic maps. In many cases, AlphaFold predictions matched experimental maps remarkably closely. In other cases, even very high-confidence predictions differed from experimental maps on a global scale through distortion and domain orientation, and on a local scale in backbone and side-chain conformation. We suggest considering AlphaFold predictions as exceptionally useful hypotheses. We further suggest that it is important to consider the confidence in prediction when interpreting AlphaFold predictions and to carry out experimental structure determination to verify structural details, particularly those that involve interactions not included in the prediction.

Improved joint X-ray and neutron refinement procedure in Phenix.

Neutron diffraction is one of the three crystallographic techniques (X-ray, neutron and electron diffraction) used to determine the atomic structures of molecules. Its particular strengths derive from the fact that H (and D) atoms are strong neutron scatterers, meaning that their positions, and thus protonation states, can be derived from crystallographic maps. However, because of technical limitations and experimental obstacles, the quality of neutron diffraction data is typically much poorer (completeness, resolution and signal to noise) than that of X-ray diffraction data for the same sample. Further, refinement is more complex as it usually requires additional parameters to describe the H (and D) atoms. The increase in the number of parameters may be mitigated by using the `riding hydrogen' refinement strategy, in which the positions of H atoms without a rotational degree of freedom are inferred from their neighboring heavy atoms. However, this does not address the issues related to poor data quality. Therefore, neutron structure determination often relies on the presence of an X-ray data set for joint X-ray and neutron (XN) refinement. In this approach, the X-ray data serve to compensate for the deficiencies of the neutron diffraction data by refining one model simultaneously against the X-ray and neutron data sets. To be applicable, it is assumed that both data sets are highly isomorphous, and preferably collected from the same crystals and at the same temperature. However, the approach has a number of limitations that are discussed in this work by comparing four separately re-refined neutron models. To address the limitations, a new method for joint XN refinement is introduced that optimizes two different models against the different data sets. This approach is tested using neutron models and data deposited in the Protein Data Bank. The efficacy of refining models with H atoms as riding or as individual atoms is also investigated.

Overall protein structure quality assessment using hydrogen-bonding parameters.

Atomic model refinement at low resolution is often a challenging task. This is mostly because the experimental data are not sufficiently detailed to be described by atomic models. To make refinement practical and ensure that a refined atomic model is geometrically meaningful, additional information needs to be used such as restraints on Ramachandran plot distributions or residue side-chain rotameric states. However, using Ramachandran plots or rotameric states as refinement targets diminishes the validating power of these tools. Therefore, finding additional model-validation criteria that are not used or are difficult to use as refinement goals is desirable. Hydrogen bonds are one of the important noncovalent interactions that shape and maintain protein structure. These interactions can be characterized by a specific geometry of hydrogen donor and acceptor atoms. Systematic analysis of these geometries performed for quality-filtered high-resolution models of proteins from the Protein Data Bank shows that they have a distinct and a conserved distribution. Here, it is demonstrated how this information can be used for atomic model validation.

Efficient structure-factor modeling for crystals with multiple components.

Diffraction intensities from a crystallographic experiment include contributions from the entire unit cell of the crystal: the macromolecule, the solvent around it and eventually other compounds. These contributions cannot typically be well described by an atomic model alone, i.e. using point scatterers. Indeed, entities such as disordered (bulk) solvent, semi-ordered solvent (e.g. lipid belts in membrane proteins, ligands, ion channels) and disordered polymer loops require other types of modeling than a collection of individual atoms. This results in the model structure factors containing multiple contributions. Most macromolecular applications assume two-component structure factors: one component arising from the atomic model and the second one describing the bulk solvent. A more accurate and detailed modeling of the disordered regions of the crystal will naturally require more than two components in the structure factors, which presents algorithmic and computational challenges. Here an efficient solution of this problem is proposed. All algorithms described in this work have been implemented in the computational crystallography toolbox (CCTBX) and are also available within Phenix software. These algorithms are rather general and do not use any assumptions about molecule type or size nor about those of its components.

Accelerating crystal structure determination with iterative AlphaFold prediction.

Experimental structure determination can be accelerated with artificial intelligence (AI)-based structure-prediction methods such as AlphaFold. Here, an automatic procedure requiring only sequence information and crystallographic data is presented that uses AlphaFold predictions to produce an electron-density map and a structural model. Iterating through cycles of structure prediction is a key element of this procedure: a predicted model rebuilt in one cycle is used as a template for prediction in the next cycle. This procedure was applied to X-ray data for 215 structures released by the Protein Data Bank in a recent six-month period. In 87% of cases our procedure yielded a model with at least 50% of Cα atoms matching those in the deposited models within 2 Å. Predictions from the iterative template-guided prediction procedure were more accurate than those obtained without templates. It is concluded that AlphaFold predictions obtained based on sequence information alone are usually accurate enough to solve the crystallographic phase problem with molecular replacement, and a general strategy for macromolecular structure determination that includes AI-based prediction both as a starting point and as a method of model optimization is suggested.

Conformational space exploration of cryo-EM structures by variability refinement.

Cryo-EM observation of biological samples enables visualization of sample heterogeneity, in the form of discrete states that are separable, or continuous heterogeneity as a result of local protein motion before flash freezing. Variability analysis of this continuous heterogeneity describes the variance between a particle stack and a volume, and results in a map series describing the various steps undertaken by the sample in the particle stack. While this observation is absolutely stunning, it is very hard to pinpoint structural details to elements of the maps. In order to bridge the gap between observation and explanation, we designed a tool that refines an ensemble of structures into all the maps from variability analysis. Using this bundle of structures, it is easy to spot variable parts of the structure, as well as the parts that are not moving. Comparison with molecular dynamics simulations highlights the fact that the movements follow the same directions, albeit with different amplitudes. Ligand can also be investigated using this method. Variability refinement is available in the Phenix software suite, accessible under the program name phenix.varref.

Improved AlphaFold modeling with implicit experimental information.

Machine-learning prediction algorithms such as AlphaFold and RoseTTAFold can create remarkably accurate protein models, but these models usually have some regions that are predicted with low confidence or poor accuracy. We hypothesized that by implicitly including new experimental information such as a density map, a greater portion of a model could be predicted accurately, and that this might synergistically improve parts of the model that were not fully addressed by either machine learning or experiment alone. An iterative procedure was developed in which AlphaFold models are automatically rebuilt on the basis of experimental density maps and the rebuilt models are used as templates in new AlphaFold predictions. We show that including experimental information improves prediction beyond the improvement obtained with simple rebuilding guided by the experimental data. This procedure for AlphaFold modeling with density has been incorporated into an automated procedure for interpretation of crystallographic and electron cryo-microscopy maps.

Protein identification from electron cryomicroscopy maps by automated model building and side-chain matching.

Using single-particle electron cryo-microscopy (cryo-EM), it is possible to obtain multiple reconstructions showing the 3D structures of proteins imaged as a mixture. Here, it is shown that automatic map interpretation based on such reconstructions can be used to create atomic models of proteins as well as to match the proteins to the correct sequences and thereby to identify them. This procedure was tested using two proteins previously identified from a mixture at resolutions of 3.2 Å, as well as using 91 deposited maps with resolutions between 2 and 4.5 Å. The approach is found to be highly effective for maps obtained at resolutions of 3.5 Šand better, and to have some utility at resolutions as low as 4 Å.

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

This paper describes outcomes of the 2019 Cryo-EM Model Challenge. The goals were to (1) assess the quality of models that can be produced from cryogenic electron microscopy (cryo-EM) maps using current modeling software, (2) evaluate reproducibility of modeling results from different software developers and users and (3) compare performance of current metrics used for model evaluation, particularly Fit-to-Map metrics, with focus on near-atomic resolution. Our findings demonstrate the relatively high accuracy and reproducibility of cryo-EM models derived by 13 participating teams from four benchmark maps, including three forming a resolution series (1.8 to 3.1 Å). The results permit specific recommendations to be made about validating near-atomic cryo-EM structures both in the context of individual experiments and structure data archives such as the Protein Data Bank. We recommend the adoption of multiple scoring parameters to provide full and objective annotation and assessment of the model, reflective of the observed cryo-EM map density.

Detection of translational noncrystallographic symmetry in Patterson functions.

Detection of translational noncrystallographic symmetry (TNCS) can be critical for success in crystallographic phasing, particularly when molecular-replacement models are poor or anomalous phasing information is weak. If the correct TNCS is detected then expected intensity factors for each reflection can be refined, so that the maximum-likelihood functions underlying molecular replacement and single-wavelength anomalous dispersion use appropriate structure-factor normalization and variance terms. Here, an analysis of a curated database of protein structures from the Protein Data Bank to investigate how TNCS manifests in the Patterson function is described. These studies informed an algorithm for the detection of TNCS, which includes a method for detecting the number of vectors involved in any commensurate modulation (the TNCS order). The algorithm generates a ranked list of possible TNCS associations in the asymmetric unit for exploration during structure solution.

CERES: a cryo-EM re-refinement system for continuous improvement of deposited models.

The field of electron cryomicroscopy (cryo-EM) has advanced quickly in recent years as the result of numerous technological and methodological developments. This has led to an increase in the number of atomic structures determined using this method. Recently, several tools for the analysis of cryo-EM data and models have been developed within the Phenix software package, such as phenix.real_space_refine for the refinement of atomic models against real-space maps. Also, new validation metrics have been developed for low-resolution cryo-EM models. To understand the quality of deposited cryo-EM structures and how they might be improved, models deposited in the Protein Data Bank that have map resolutions of better than 5 Šwere automatically re-refined using current versions of Phenix tools. The results are available on a publicly accessible web page (https://cci.lbl.gov/ceres). The implementation of a Cryo-EM Re-refinement System (CERES) for the improvement of models deposited in the wwPDB, and the results of the re-refinements, are described. Based on these results, contents are proposed for a `cryo-EM Table 1', which summarizes experimental details and validation metrics in a similar way to `Table 1' in crystallography. The consistent use of robust metrics for the evaluation of cryo-EM models and data should accompany every structure deposition and be reported in scientific publications.

Real-space quantum-based refinement for cryo-EM: Q|R#3.

Electron cryo-microscopy (cryo-EM) is rapidly becoming a major competitor to X-ray crystallography, especially for large structures that are difficult or impossible to crystallize. While recent spectacular technological improvements have led to significantly higher resolution three-dimensional reconstructions, the average quality of cryo-EM maps is still at the low-resolution end of the range compared with crystallography. A long-standing challenge for atomic model refinement has been the production of stereochemically meaningful models for this resolution regime. Here, it is demonstrated that including accurate model geometry restraints derived from ab initio quantum-chemical calculations (HF-D3/6-31G) can improve the refinement of an example structure (chain A of PDB entry 3j63). The robustness of the procedure is tested for additional structures with up to 7000 atoms (PDB entry 3a5x and chain C of PDB entry 5fn5) using the less expensive semi-empirical (GFN1-xTB) model. The necessary algorithms enabling real-space quantum refinement have been implemented in the latest version of qr.refine and are described here.

Density modification of cryo-EM maps.

Density modification uses expectations about features of a map such as a flat solvent and expected distributions of density in the region of the macromolecule to improve individual Fourier terms representing the map. This process transfers information from one part of a map to another and can improve the accuracy of a map. Here, the assumptions behind density modification for maps from electron cryomicroscopy are examined and a procedure is presented that allows the incorporation of model-based information. Density modification works best in cases where unfiltered, unmasked maps with clear boundaries between the macromolecule and solvent are visible, and where there is substantial noise in the map, both in the region of the macromolecule and the solvent. It also is most effective if the characteristics of the map are relatively constant within regions of the macromolecule and the solvent. Model-based information can be used to improve density modification, but model bias can in principle occur. Here, model bias is reduced by using ensemble models that allow an estimation of model uncertainty. A test of model bias is presented that suggests that even if the expected density in a region of a map is specified incorrectly by using an incorrect model, the incorrect expectations do not strongly affect the final map.

Improvement of cryo-EM maps by density modification.

A density-modification procedure for improving maps from single-particle electron cryogenic microscopy (cryo-EM) is presented. The theoretical basis of the method is identical to that of maximum-likelihood density modification, previously used to improve maps from macromolecular X-ray crystallography. Key differences from applications in crystallography are that the errors in Fourier coefficients are largely in the phases in crystallography but in both phases and amplitudes in cryo-EM, and that half-maps with independent errors are available in cryo-EM. These differences lead to a distinct approach for combination of information from starting maps with information obtained in the density-modification process. The density-modification procedure was applied to a set of 104 datasets and improved map-model correlation and increased the visibility of details in many of the maps. The procedure requires two unmasked half-maps and a sequence file or other source of information on the volume of the macromolecule that has been imaged.

A Global Ramachandran Score Identifies Protein Structures with Unlikely Stereochemistry.

Ramachandran plots report the distribution of the (ϕ, ψ) torsion angles of the protein backbone and are one of the best quality metrics of experimental structure models. Typically, validation software reports the number of residues belonging to "outlier," "allowed," and "favored" regions. While "zero unexplained outliers" can be considered the current "gold standard," this can be misleading if deviations from expected distributions are not considered. We revisited the Ramachandran Z score (Rama-Z), a quality metric introduced more than two decades ago but underutilized. We describe a reimplementation of the Rama-Z score in the Computational Crystallography Toolbox along with an algorithm to estimate its uncertainty for individual models; final implementations are available in Phenix and PDB-REDO. We discuss the interpretation of the Rama-Z score and advocate including it in the validation reports provided by the Protein Data Bank. We also advocate reporting it alongside the outlier/allowed/favored counts in structural publications.

Implementation of the riding hydrogen model in CCTBX to support the next generation of X-ray and neutron joint refinement in Phenix.

A fundamental prerequisite for implementing new procedures of atomic model refinement against neutron diffraction data is the efficient handling of hydrogen atoms. The riding hydrogen model, which constrains hydrogen atom parameters to those of the non-hydrogen atoms, is a plausible parameterization for refinements. This work describes the implementation of the riding hydrogen model in the Computational Crystallography Toolbox and in Phenix. Riding hydrogen atoms can be found in several different configurations that are characterized by specific geometries. For each configuration, the hydrogen atom parameterization and the expressions for the gradients of refinement target function with respect to non-hydrogen parameters are described.

What are the current limits on determination of protonation state using neutron macromolecular crystallography?

The rate of deposition of models determined by neutron diffraction, or a hybrid approach that combines X-ray and neutron diffraction, has increased in recent years. The benefit of neutron diffraction is that hydrogen atom (H) positions are detectable, allowing for the determination of protonation state and water molecule orientation. This study analyses all neutron models deposited in the Protein Data Bank to date, focusing on protonation state and properties of H (or deuterium, D) atoms as well as the details of water molecules. In particular, clashes and hydrogen bonds involving H or D atoms are investigated. As water molecules are typically the least reproducible part of a structural model, their positions in neutron models were compared to those in homologous high-resolution X-ray structures. For models determined by joint refinement against X-ray and neutron data, the water structure comparison was also carried out for models re-refined against the X-ray data alone. The homologues have generally fewer conserved water molecules where X-ray only was used and the positions of equivalent waters vary more than in the case of the hybrid X-ray model. As neutron diffraction data are generally less complete than X-ray data, the influence of neutron data completeness on nuclear density maps was also analyzed. We observe and discuss systematic map quality deterioration as result of data incompleteness.