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.

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.

In situ ligand restraints from quantum-mechanical methods.

In macromolecular crystallographic structure refinement, ligands present challenges for the generation of geometric restraints due to their large chemical variability, their possible novel nature and their specific interaction with the binding pocket of the protein. Quantum-mechanical approaches are useful for providing accurate ligand geometries, but can be plagued by the number of minima in flexible molecules. In an effort to avoid these issues, the Quantum Mechanical Restraints (QMR) procedure optimizes the ligand geometry in situ, thus accounting for the influence of the macromolecule on the local energy minima of the ligand. The optimized ligand geometry is used to generate target values for geometric restraints during the crystallographic refinement. As demonstrated using a sample of >2330 ligand instances in >1700 protein-ligand models, QMR restraints generally result in lower deviations from the target stereochemistry compared with conventionally generated restraints. In particular, the QMR approach provides accurate torsion restraints for ligands and other entities.

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.

Arginine off-kilter: guanidinium is not as planar as restraints denote.

Crystallographic refinement of macromolecular structures relies on stereochemical restraints to mitigate the typically poor data-to-parameter ratio. For proteins, each amino acid has a unique set of geometry restraints which represent stereochemical information such as bond lengths, valence angles, torsion angles, dihedrals and planes. It has been shown that the geometry in refined structures can differ significantly from that present in libraries; for example, it was recently reported that the guanidinium moiety in arginine is not symmetric. In this work, the asymmetry of the Nϵ-Cζ-Nη1 and Nϵ-Cζ-Nη2 valence angles in the guanidinium moiety is confirmed. In addition, it was found that the Cδ atom can deviate significantly (more than 20°) from the guanidinium plane. This requires the relaxation of the planar restraint for the Cδ atom, as it otherwise causes the other atoms in the group to compensate by distorting the guanidinium core plane. A new set of restraints for the arginine side chain have therefore been formulated, and are available in the software package Phenix, that take into account the asymmetry of the group and the planar deviation of the Cδ atom. This is an example of the need to regularly revisit the geometric restraint libraries used in macromolecular refinement so that they reflect the best knowledge of the structural chemistry of their components available at the time.

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.

Macromolecular structure determination using X-rays, neutrons and electrons: recent developments in Phenix.

Diffraction (X-ray, neutron and electron) and electron cryo-microscopy are powerful methods to determine three-dimensional macromolecular structures, which are required to understand biological processes and to develop new therapeutics against diseases. The overall structure-solution workflow is similar for these techniques, but nuances exist because the properties of the reduced experimental data are different. Software tools for structure determination should therefore be tailored for each method. Phenix is a comprehensive software package for macromolecular structure determination that handles data from any of these techniques. Tasks performed with Phenix include data-quality assessment, map improvement, model building, the validation/rebuilding/refinement cycle and deposition. Each tool caters to the type of experimental data. The design of Phenix emphasizes the automation of procedures, where possible, to minimize repetitive and time-consuming manual tasks, while default parameters are chosen to encourage best practice. A graphical user interface provides access to many command-line features of Phenix and streamlines the transition between programs, project tracking and re-running of previous tasks.

Evaluation of models determined by neutron diffraction and proposed improvements to their validation and deposition.

The Protein Data Bank (PDB) contains a growing number of models that have been determined using neutron diffraction or a hybrid method that combines X-ray and neutron diffraction. The advantage of neutron diffraction experiments is that the positions of all atoms can be determined, including H atoms, which are hardly detectable by X-ray diffraction. This allows the determination of protonation states and the assignment of H atoms to water molecules. Because neutrons are scattered differently by hydrogen and its isotope deuterium, neutron diffraction in combination with H/D exchange can provide information on accessibility, dynamics and chemical lability. In this study, the deposited data, models and model-to-data fit for all PDB entries that used neutron diffraction as the source of experimental data have been analysed. In many cases, the reported Rwork and Rfree values were not reproducible. In such cases, the model and data files were analysed to identify the reasons for this mismatch. The issues responsible for the discrepancies are summarized and explained. The analysis unveiled limitations to the annotation, deposition and validation of models and data, and a lack of community-wide accepted standards for the description of neutron models and data, as well as deficiencies in current model refinement tools. Most of the issues identified concern the handling of H atoms. Since the primary use of neutron macromolecular crystallography is to locate and directly visualize H atoms, it is important to address these issues, so that the deposited neutron models allow the retrieval of the maximum amount of information with the smallest effort of manual intervention. A path forward to improving the annotation, validation and deposition of neutron models and hybrid X-ray and neutron models is suggested.

Interactive comparison and remediation of collections of macromolecular structures.

Often similar structures need to be compared to reveal local differences throughout the entire model or between related copies within the model. Therefore, a program to compare multiple structures and enable correction any differences not supported by the density map was written within the Phenix framework (Adams et al., Acta Cryst 2010; D66:213-221). This program, called Structure Comparison, can also be used for structures with multiple copies of the same protein chain in the asymmetric unit, that is, as a result of non-crystallographic symmetry (NCS). Structure Comparison was designed to interface with Coot(Emsley et al., Acta Cryst 2010; D66:486-501) and PyMOL(DeLano, PyMOL 0.99; 2002) to facilitate comparison of large numbers of related structures. Structure Comparison analyzes collections of protein structures using several metrics, such as the rotamer conformation of equivalent residues, displays the results in tabular form and allows superimposed protein chains and density maps to be quickly inspected and edited (via the tools in Coot) for consistency, completeness and correctness.

The first crystal structure of a family 129 glycoside hydrolase from a probiotic bacterium reveals critical residues and metal cofactors.

The α-N-acetylgalactosaminidase from the probiotic bacterium Bifidobacterium bifidum (NagBb) belongs to the glycoside hydrolase family 129 and hydrolyzes the glycosidic bond of Tn-antigen (GalNAcα1-Ser/Thr). NagBb is involved in assimilation of O-glycans on mucin glycoproteins by B. bifidum in the human gastrointestinal tract, but its catalytic mechanism has remained elusive because of a lack of sequence homology around putative catalytic residues and of other structural information. Here we report the X-ray crystal structure of NagBb, representing the first GH129 family structure, solved by the single-wavelength anomalous dispersion method based on sulfur atoms of the native protein. We determined ligand-free, GalNAc, and inhibitor complex forms of NagBb and found that Asp-435 and Glu-478 are located in the catalytic domain at appropriate positions for direct nucleophilic attack at the anomeric carbon and proton donation for the glycosidic bond oxygen, respectively. A highly conserved Asp-330 forms a hydrogen bond with the O4 hydroxyl of GalNAc in the -1 subsite, and Trp-398 provides a stacking platform for the GalNAc pyranose ring. Interestingly, a metal ion, presumably Ca2+, is involved in the recognition of the GalNAc N-acetyl group. Mutations at Asp-435, Glu-478, Asp-330, and Trp-398 and residues involved in metal coordination (including an all-Ala quadruple mutant) significantly reduced the activity, indicating that these residues and the metal ion play important roles in substrate recognition and catalysis. Interestingly, NagBb exhibited some structural similarities to the GH101 endo-α-N-acetylgalactosaminidases, but several critical differences in substrate recognition and reaction mechanism account for the different activities of these two enzymes.

Polder maps: improving OMIT maps by excluding bulk solvent.

The crystallographic maps that are routinely used during the structure-solution workflow are almost always model-biased because model information is used for their calculation. As these maps are also used to validate the atomic models that result from model building and refinement, this constitutes an immediate problem: anything added to the model will manifest itself in the map and thus hinder the validation. OMIT maps are a common tool to verify the presence of atoms in the model. The simplest way to compute an OMIT map is to exclude the atoms in question from the structure, update the corresponding structure factors and compute a residual map. It is then expected that if these atoms are present in the crystal structure, the electron density for the omitted atoms will be seen as positive features in this map. This, however, is complicated by the flat bulk-solvent model which is almost universally used in modern crystallographic refinement programs. This model postulates constant electron density at any voxel of the unit-cell volume that is not occupied by the atomic model. Consequently, if the density arising from the omitted atoms is weak then the bulk-solvent model may obscure it further. A possible solution to this problem is to prevent bulk solvent from entering the selected OMIT regions, which may improve the interpretative power of residual maps. This approach is called a polder (OMIT) map. Polder OMIT maps can be particularly useful for displaying weak densities of ligands, solvent molecules, side chains, alternative conformations and residues both in terminal regions and in loops. The tools described in this manuscript have been implemented and are available in PHENIX.

Crystallographic analysis of murine p24γ2 Golgi dynamics domain.

The p24 family proteins form homo- and hetero-oligomeric complexes for efficient transport of cargo proteins from the endoplasmic reticulum to the Golgi apparatus. It consists of four subfamilies (p24α, p24ß, p24γ, and p24δ). p24γ2 plays crucial roles in the selective transport of glycosylphosphatidylinositol-anchored proteins. Here, we determined the crystal structure of mouse p24γ2 Golgi dynamics (GOLD) domain at 2.8 Å resolution by the single anomalous diffraction method using intrinsic sulfur atoms. In spite of low sequence identity among p24 family proteins, p24γ2 GOLD domain assumes a ß-sandwich fold, similar to that of p24ß1 or p24δ1. An additional short α-helix is observed at the C-terminus of the p24γ2 GOLD domain. Intriguingly, p24γ2 GOLD domains crystallize as dimers, and dimer formation seems assisted by the short α-helix. Dimerization modes of GOLD domains are compared among p24 family proteins. Proteins 2017; 85:764-770. © 2016 Wiley Periodicals, Inc.

On the influence of crystal size and wavelength on native SAD phasing.

Native SAD is an emerging phasing technique that uses the anomalous signal of native heavy atoms to obtain crystallographic phases. The method does not require specific sample preparation to add anomalous scatterers, as the light atoms contained in the native sample are used as marker atoms. The most abundant anomalous scatterer used for native SAD, which is present in almost all proteins, is sulfur. However, the absorption edge of sulfur is at low energy (2.472 keV = 5.016 Å), which makes it challenging to carry out native SAD phasing experiments as most synchrotron beamlines are optimized for shorter wavelength ranges where the anomalous signal of sulfur is weak; for longer wavelengths, which produce larger anomalous differences, the absorption of X-rays by the sample, solvent, loop and surrounding medium (e.g. air) increases tremendously. Therefore, a compromise has to be found between measuring strong anomalous signal and minimizing absorption. It was thus hypothesized that shorter wavelengths should be used for large crystals and longer wavelengths for small crystals, but no thorough experimental analyses have been reported to date. To study the influence of crystal size and wavelength, native SAD experiments were carried out at different wavelengths (1.9 and 2.7 Šwith a helium cone; 3.0 and 3.3 Šwith a helium chamber) using lysozyme and ferredoxin reductase crystals of various sizes. For the tested crystals, the results suggest that larger sample sizes do not have a detrimental effect on native SAD data and that long wavelengths give a clear advantage with small samples compared with short wavelengths. The resolution dependency of substructure determination was analyzed and showed that high-symmetry crystals with small unit cells require higher resolution for the successful placement of heavy atoms.

Radiation decay of thaumatin crystals at three X-ray energies.

Radiation damage is an unavoidable obstacle in X-ray crystallographic data collection for macromolecular structure determination, so it is important to know how much radiation a sample can endure before being degraded beyond an acceptable limit. In the literature, the threshold at which the average intensity of all recorded reflections decreases to a certain fraction of the initial value is called the `dose limit'. The first estimated D50 dose-limit value, at which the average diffracted intensity was reduced to 50%, was 20 MGy and was derived from observing sample decay in electron-diffraction experiments. A later X-ray study carried out at 100 K on ferritin protein crystals arrived at a D50 of 43 MGy, and recommended an intensity reduction of protein reflections to 70%, D70, corresponding to an absorbed dose of 30 MGy, as a more appropriate limit for macromolecular crystallography. In the macromolecular crystallography community, the rate of intensity decay with dose was then assumed to be similar for all protein crystals. A series of diffraction images of cryocooled (100 K) thaumatin crystals at identical small, 2° rotation intervals were recorded at X-ray energies of 6.33 , 12.66 and 19.00 keV. Five crystals were used for each wavelength. The decay in the average diffraction intensity to 70% of the initial value, for data extending to 2.45 Šresolution, was determined to be about 7.5 MGy at 6.33 keV and about 11 MGy at the two higher energies.

Hydrogen atoms in protein structures: high-resolution X-ray diffraction structure of the DFPase.

BACKGROUND: Hydrogen atoms represent about half of the total number of atoms in proteins and are often involved in substrate recognition and catalysis. Unfortunately, X-ray protein crystallography at usual resolution fails to access directly their positioning, mainly because light atoms display weak contributions to diffraction. However, sub-Ångstrom diffraction data, careful modeling and a proper refinement strategy can allow the positioning of a significant part of hydrogen atoms. RESULTS: A comprehensive study on the X-ray structure of the diisopropyl-fluorophosphatase (DFPase) was performed, and the hydrogen atoms were modeled, including those of solvent molecules. This model was compared to the available neutron structure of DFPase, and differences in the protein and the active site solvation were noticed. CONCLUSIONS: A further examination of the DFPase X-ray structure provides substantial evidence about the presence of an activated water molecule that may constitute an interesting piece of information as regard to the enzymatic hydrolysis mechanism.

On the reproducibility of protein crystal structures: five atomic resolution structures of trypsin.

Structural studies of proteins usually rely on a model obtained from one crystal. By investigating the details of this model, crystallographers seek to obtain insight into the function of the macromolecule. It is therefore important to know which details of a protein structure are reproducible or to what extent they might differ. To address this question, the high-resolution structures of five crystals of bovine trypsin obtained under analogous conditions were compared. Global parameters and structural details were investigated. All of the models were of similar quality and the pairwise merged intensities had large correlation coefficients. The C(α) and backbone atoms of the structures superposed very well. The occupancy of ligands in regions of low thermal motion was reproducible, whereas solvent molecules containing heavier atoms (such as sulfur) or those located on the surface could differ significantly. The coordination lengths of the calcium ion were conserved. A large proportion of the multiple conformations refined to similar occupancies and the residues adopted similar orientations. More than three quarters of the water-molecule sites were conserved within 0.5 Šand more than one third were conserved within 0.1 Å. An investigation of the protonation states of histidine residues and carboxylate moieties was consistent for all of the models. Radiation-damage effects to disulfide bridges were observed for the same residues and to similar extents. Main-chain bond lengths and angles averaged to similar values and were in agreement with the Engh and Huber targets. Other features, such as peptide flips and the double conformation of the inhibitor molecule, were also reproducible in all of the trypsin structures. Therefore, many details are similar in models obtained from different crystals. However, several features of residues or ligands located in flexible parts of the macromolecule may vary significantly, such as side-chain orientations and the occupancies of certain fragments.