This article was generated entirely by AI without human authorship. Please read with discretion, or have an AI verify its accuracy.
AI TranslationSimplified ChineseEnglish
Key Insights
This issue includes 51 papers covering cryo-EM reconstruction and validation, atomic models and chemical checks, crystallographic data quality, AlphaFold predictions, protein dynamics and allostery, membrane receptors, neuroscience, structural alignment, and biological assemblies. The central emphasis is on validation using multiple metrics, support from local data, and independent experimental verification: global resolution, a good fit, or high prediction confidence alone cannot guarantee that structural details are correct. Studies of conformational ensembles and dynamics complement interpretations based on static structures. New technologies expand capabilities in sample preparation, in situ observation, and drug design, but specific conclusions remain constrained by the samples, experimental conditions, and scope of evaluation. Entries without abstracts list metadata only; no content is inferred.
This issue features 51 papers spanning cryo-EM reconstruction and validation, atomic models and chemical validation, crystallographic data quality, AlphaFold structure prediction, protein dynamics and allostery, membrane receptors and ion channels, neuroscience, structural alignment, and biological assemblies.
Cryo-EM reconstruction, resolution, and validation (11 papers)
01 · RELION: Implementation of a Bayesian approach to cryo-EM structure determination
Journal of Structural Biology · Publication / issue date: 2012
Authors: Sjors H.W. Scheres
Single-particle cryo-EM reconstruction often involves many parameters that must be tuned by experience. RELION frames this problem in a Bayesian statistical framework. The software estimates statistical model parameters from the data and refines macromolecular structures using maximum a posteriori probability, reducing the need for repeated manual tuning. The paper describes an implementation that lowers computational costs and examines how precisely the relative orientations of individual particles can be determined. It also incorporates gold-standard Fourier shell correlation to guard against overfitting. The authors report that the implementation can produce high-quality reconstructions and reliable resolution estimates at acceptable computational cost, with limited user intervention. Its key contribution is to place parameter estimation, reconstruction, and validation within explicit statistical assumptions. In practice, results still depend on the input data and optimization conditions, so automation should not be mistaken for a substitute for inspection.
Highlight · RELION combines Bayesian parameter estimation with independent half-data validation, enabling single-particle reconstruction with less manual tuning.
Authors: Pavel V. Afonine, Bruno P. Klaholz, Nigel W. Moriarty, Billy K. Poon, Oleg V. Sobolev, Thomas C. Terwilliger, Paul D. Adams, Alexandre Urzhumtsev
As the number of atomic models derived from cryo-EM has grown rapidly, simply reusing crystallographic checks has proved insufficient to capture the specific characteristics of maps. This paper presents tools in PHENIX for cryo-EM analysis and validation, including d99 for estimating resolution, phenix.auto_sharpen for improving map appearance, and phenix.mtriage for analyzing map properties. It also demonstrates their use on structures and maps deposited in public archives. The authors recommend retaining half-maps and masks so that outside users can more fully assess reconstructions and atomic models. The work connects map processing and resolution description with model validation, and underscores the importance of making data available for review. The abstract does not prescribe a fixed set of processing parameters for every sample, so these tools should be treated as complementary diagnostics and used in light of the original data and local features.
Highlight · Cryo-EM map tools, together with archived half-maps and masks, support structural validation that others can independently review.
03 · Outcomes of the EMDataResource cryo-EM Ligand Modeling Challenge
Nature Methods · Publication / issue date: 2024
Authors: Catherine L. Lawson, Andriy Kryshtafovych, Grigore D. Pintilie, Stephen K. Burley, Jiří Černý, Vincent B. Chen, Paul Emsley, Alberto Gobbi, Andrzej Joachimiak, Sigrid Noreng, Michael G. Prisant, Randy J. Read, Jane S. Richardson, Alexis L. Rohou, Bohdan Schneider, Benjamin D. Sellers, Chenghua Shao, Elizabeth Sourial, Chris I. Williams, Christopher J. Williams, Ying Yang, Venkat Abbaraju, Pavel V. Afonine, Matthew L. Baker, Paul S. Bond, Tom L. Blundell, Tom Burnley, Arthur Campbell, Renzhi Cao, Jianlin Cheng, Grzegorz Chojnowski, K. D. Cowtan, Frank DiMaio, Reza Esmaeeli, Nabin Giri, Helmut Grubmüller, Soon Wen Hoh, Jie Hou, Corey F. Hryc, Carola Hunte, Maxim Igaev, Agnel P. Joseph, Wei-Chun Kao, Daisuke Kihara, Dilip Kumar, Lijun Lang, Sean Lin, Sai R. Maddhuri Venkata Subramaniya, Sumit Mittal, Arup Mondal, Nigel W. Moriarty, Andrew Muenks, Garib N. Murshudov, Robert A. Nicholls, Mateusz Olek, Colin M. Palmer, Alberto Perez, Emmi Pohjolainen, Karunakar R. Pothula, Christopher N. Rowley, Daipayan Sarkar, Luisa U. Schäfer, Christopher J. Schlicksup, Gunnar F. Schröder, Mrinal Shekhar, Dong Si, Abhishek Singharoy, Oleg V. Sobolev, Genki Terashi, Andrea C. Vaiana, Sundeep C. Vedithi, Jacob Verburgt, Xiao Wang, Rangana Warshamanage, Martyn D. Winn, Simone Weyand, Keitaro Yamashita, Minglei Zhao, Michael F. Schmid, Helen M. Berman, Wah Chiu
A near-atomic-resolution map may reveal a ligand, but that does not mean different modelers will arrive at equally reliable binding poses. The EMDataResource ligand-modeling challenge used three maps at 1.9–2.5 Å resolution, covering inhibitor, covalent nucleotide analog, and lipid-binding cases. Seventeen independent teams submitted 61 models and workflows. The study assessed ligand and surrounding-atom modeling using visual inspection, local map quality, map fit, geometry, energy, and contact scores. The results show that no single score can capture the overall quality of a macromolecule and its ligand; multiple metrics need to be considered together. The paper therefore proposes practices for evaluating ligand structures at near-atomic resolution, emphasizing that neither global resolution nor a good fit score alone guarantees a correct local chemical interpretation.
Highlight · Ligand models should be evaluated using multiple lines of evidence, including map fit, geometry, energy, and contacts.
04 · Cryo-EM model validation recommendations based on outcomes of the 2019 EMDataResource challenge
Nature Methods · Publication / issue date: 2021
Authors: Catherine L. Lawson, Andriy Kryshtafovych, Paul D. Adams, Pavel V. Afonine, Matthew L. Baker, Benjamin A. Barad, Paul Bond, Tom Burnley, Renzhi Cao, Jianlin Cheng, Grzegorz Chojnowski, Kevin Cowtan, Ken A. Dill, Frank DiMaio, Daniel P. Farrell, James S. Fraser, Mark A. Herzik, Soon Wen Hoh, Jie Hou, Li-Wei Hung, Maxim Igaev, Agnel P. Joseph, Daisuke Kihara, Dilip Kumar, Sumit Mittal, Bohdan Monastyrskyy, Mateusz Olek, Colin M. Palmer, Ardan Patwardhan, Alberto Perez, Jonas Pfab, Grigore D. Pintilie, Jane S. Richardson, Peter B. Rosenthal, Daipayan Sarkar, Luisa U. Schäfer, Michael F. Schmid, Gunnar F. Schröder, Mrinal Shekhar, Dong Si, Abishek Singharoy, Genki Terashi, Thomas C. Terwilliger, Andrea Vaiana, Liguo Wang, Zhe Wang, Stephanie A. Wankowicz, Christopher J. Williams, Martyn Winn, Tianqi Wu, Xiaodi Yu, Kaiming Zhang, Helen M. Berman, Wah Chiu
The 2019 cryo-EM model challenge set out to assess the quality of models produced by existing software, how reproducible results were across teams, and how well validation metrics—particularly map-fitting metrics—performed. Thirteen teams built models for four benchmark maps, three of which formed a resolution series spanning 1.8–3.1 Å. The comparison found that models at near-atomic resolution were generally accurate and reproducible, while different scores captured different aspects of quality. The authors therefore recommend using multiple metrics to provide a comprehensive, objective description of models and how well they fit the observed density. These recommendations apply both to individual experiments and to structural archives such as the PDB. Because the findings are based on the benchmark maps used, achieving the same nominal resolution does not by itself mean that every sample will yield an equally reliable model.
Highlight · The community challenge supports validation with multiple metrics, capturing both reproducibility and support from the experimental map.
05 · Community recommendations on cryoEM data archiving and validation
IUCrJ · Publication / issue date: 2024
Authors: Gerard J. Kleywegt, Paul D. Adams, Sarah J. Butcher, Catherine L. Lawson, Alexis Rohou, Peter B. Rosenthal, Sriram Subramaniam, Maya Topf, Sanja Abbott, Philip R. Baldwin, John M. Berrisford, Gérard Bricogne, Preeti Choudhary, Tristan I. Croll, Radostin Danev, Sai J. Ganesan, Timothy Grant, Aleksandras Gutmanas, Richard Henderson, J. Bernard Heymann, Juha T. Huiskonen, Andrei Istrate, Takayuki Kato, Gabriel C. Lander, Shee-Mei Lok, Steven J. Ludtke, Garib N. Murshudov, Ryan Pye, Grigore D. Pintilie, Jane S. Richardson, Carsten Sachse, Osman Salih, Sjors H. W. Scheres, Gunnar F. Schroeder, Carlos Oscar S. Sorzano, Scott M. Stagg, Zhe Wang, Rangana Warshamanage, John D. Westbrook, Martyn D. Winn, Jasmine Y. Young, Stephen K. Burley, Jeffrey C. Hoch, Genji Kurisu, Kyle Morris, Ardan Patwardhan, Sameer Velankar
Whether cryo-EM structures can be independently reviewed depends on what data are deposited and how maps and models are validated. This paper summarizes a 2020 workshop held at EMBL-EBI, where 47 experts in data processing, model building and refinement, validation, and archiving discussed these issues, with a focus on single-particle analysis. The report describes the workshop background and topics, the consensus recommendations reached, progress in implementing some of them, and challenges for future method development. It is a report on community consensus and implementation progress, not a new algorithm or a single experiment. Since the abstract does not list the archiving requirements item by item, we do not add a checklist here; consult the paper for specific deposition and validation details, and follow the requirements of the relevant database.
Highlight · Community consensus links cryo-EM data archiving with structural validation, making results easier to review independently.
06 · Measurement of atom resolvability in cryo-EM maps with Q-scores
Nature Methods · Publication / issue date: 2020
Authors: Grigore Pintilie, Kaiming Zhang, Zhaoming Su, Shanshan Li, Michael F. Schmid, Wah Chiu
Atoms are not equally resolvable throughout a cryo-EM map, so a single global resolution value cannot capture these differences. The study introduced Q-score, which uses a fitted or map-derived model to quantify how well individual atoms are resolved in the map. It applies to proteins, nucleic acids, water, ligands, and other solvent atoms. Scores can be summarized by residue, nucleotide, or across an entire model; the overall mean is strongly correlated with estimated resolutions for protein and RNA maps. The authors also compared maps of the same protein collected by different laboratories to test how reproducible features from side chains to water and ions are. The metric has one clear prerequisite: the model used must fit the map well. Q-score therefore provides local evidence, but cannot be treated as a fully independent measure of map quality without considering the model's placement.
Highlight · Q-score extends resolvability assessment from the map as a whole to individual atoms, but its interpretation depends on how well the model fits the map.
Journal of Structural Biology · Publication date: 2005
Authors: Marin van Heel, Michael Schatz
Fourier shell correlation (FSC) is a widely used quality metric for 3D electron microscopy reconstructions, but converting the curve into a resolution value requires choosing a threshold. This paper examines how the number of voxels per shell, structural symmetry, and the fraction of the reconstruction volume occupied by the structure affect FSC statistics. Through theory and model experiments, the authors question the practice of applying a single fixed threshold universally. They compare two types of threshold curves: those based on the sigma factor ask whether the signal is significantly above noise, while those based on bit information ask whether enough interpretable information has been collected. The paper lays bare the statistical assumptions behind resolution values and stresses that thresholds must have a clear meaning. These are the framework and arguments advanced in the paper; they cannot be detached from the specific estimation procedure and used as a replacement for validation standards in all subsequent reconstruction methods.
Highlight · The meaning of an FSC threshold depends on its statistical assumptions: detecting signal and obtaining enough interpretable information are not the same criterion.
08 · Optimal Determination of Particle Orientation, Absolute Hand, and Contrast Loss in Single-particle Electron Cryomicroscopy
Journal of Molecular Biology · Publication date: 2003
Authors: Peter B. Rosenthal, Richard Henderson
Single-particle electron microscopy requires not only determining particle orientations but also establishing the absolute hand of the 3D map. The paper uses images of the same particle field taken at two tilt angles, comparing the second image with projections of the map under the known tilt relationship and its inverse to distinguish the correct hand from the incorrect one. The authors also use this independent geometric constraint to optimize orientation searches and refinement, reducing the tendency of a model to match noise in a single image. They validated the method on image pairs of the E2 core of the pyruvate dehydrogenase complex, and recovered high-resolution contrast using noise-correlated structure-factor weighting and a temperature factor. The work links hand determination, orientation reliability, and map contrast; its resolution results come from specific samples and data and should not be taken as a general performance guarantee.
Highlight · Tilt-pair images provide an independent geometric constraint that can check map handedness and help optimize particle orientations.
09 · High-resolution noise substitution to measure overfitting and validate resolution in 3D structure determination by single particle electron cryomicroscopy
Ultramicroscopy · Publication date: 2013
Authors: Shaoxia Chen, Greg McMullan, Abdul R. Faruqi, Garib N. Murshudov, Judith M. Short, Sjors H.W. Scheres, Richard Henderson
Iterative single-particle reconstruction can amplify high-frequency noise into apparently reproducible detail, leading to an overestimate of resolution. This study proposes a high-resolution noise-substitution test: Fourier components beyond a specified resolution are replaced with background noise, or their phases are randomized while preserving the corresponding power distribution, and the data are then reconstructed using the same workflow. Comparing FSC curves for the original and noise-substituted data estimates the correlations introduced by overfitting in the final map, allowing resolution estimates to be corrected. The authors tested the method on two sets of β-galactosidase images and with several processing packages, finding that the degree of overfitting depends substantially on the processing workflow. A key advantage is that the method is software-independent and can serve as a parallel control, rather than relying on a more attractive map or a higher FSC alone as evidence of genuine added detail.
Highlight · High-frequency noise-substitution controls can quantify overfitting in reconstruction and provide an independent check on resolution estimates.
10 · Prevention of overfitting in cryo-EM structure determination
Nature Methods · Publication date: 2012
Authors: Sjors H W Scheres, Shaoxia Chen
No abstract was saved in Zotero, and neither the publisher's page nor PubMed provides an accessible abstract. This issue lists metadata only and does not infer the paper's validation workflow from its title.
11 · Quantifying the local resolution of cryo-EM density maps
Nature Methods · Publication date: 2014
Authors: Alp Kucukelbir, Fred J Sigworth, Hemant D Tagare
A single global resolution value cannot capture differences in information across regions of a cryo-EM map. The study defines local resolution in a 3D map using local sinusoidal features and implements the method in ResMap; the abstract states that the algorithm requires no free parameters and can also be applied to other imaging methods, including tomography. The authors analyzed four single-particle reconstructions and sub-tomogram averages, finding that local resolution can range from about 4 to 40 Å. The results highlight that a map can contain both well-resolved and blurred regions, and that the structural information it supports should be interpreted at the relevant scale. The abstract does not specify accuracy limits for all samples, so a particular local estimate should not be taken as direct proof that an atomic model is correct; it is better used to inform local interpretation and further checks.
Highlight · ResMap uses local features to describe variation in map resolution, revealing spatial differences that a global value cannot show.
Atomic models, sequence assignment, and chemical validation (7 papers)
12 · Sequence-assignment validation in cryo-EM models with checkMySequence
Acta Crystallographica Section D Structural Biology · Publication date: 2022
Author: Grzegorz Chojnowski
AI-based structure prediction tools have made cryo-EM modeling easier, but map interpretation, local rebuilding, and refinement can still introduce errors. Register shifts in sequence assignment are especially difficult to spot. This paper introduces checkMySequence, a fast, fully automated method that requires no parameter tuning and identifies mismatches between the residue sequence in a cryo-EM model and its position in the density. The authors show how it can aid modeling when local resolution is too low for visual inspection, and use a much-discussed sequence register error in an RNA polymerase model to illustrate the value of automated checks. The tool targets one specific source of error—sequence register—and complements conventional geometry and map-fit checks. The abstract does not claim that it detects every model problem, so it is best included in a workflow with multiple validation methods rather than used as a single quality criterion.
Highlight · checkMySequence automates the detection of sequence register shifts that are difficult to spot by eye.
13 · Using deep-learning predictions reveals a large number of register errors in PDB depositions
IUCrJ · Publication date: 2024
Authors: Filomeno Sánchez Rodríguez, Adam J. Simpkin, Grzegorz Chojnowski, Ronan M. Keegan, Daniel J. Rigden
Sequence register errors in structural models can escape routine geometry checks and may not noticeably impair map fit. The authors compare residue contacts and distances in experimental models with relationships predicted by methods such as AlphaFold2, flagging regions where the two disagree. This approach is independent of stereochemical and map-model consistency checks and does not directly depend on experimental resolution. A scan of 3–5 Å structures in the PDB identified thousands of possible register errors and proposed candidate corrections; limited implementation of these corrections improved refinement statistics in most cases. The paper also discusses confounding factors, including fold-switching proteins. Disagreement with predictions can therefore prompt a closer look at a model, but a “possible error” still needs to be assessed against the data and does not establish that every flagged region is wrong.
Highlight · Predicted contact relationships provide evidence independent of geometry and map fit, helping locate and correct sequence register errors.
14 · CheckMyMetal: a macromolecular metal-binding validation tool
Acta Crystallographica Section D Structural Biology · Publication date: 2017
Authors: Heping Zheng, David R. Cooper, Przemyslaw J. Porebski, Ivan G. Shabalin, Katarzyna B. Handing, Wladek Minor
Metal ions participate in many biological processes, but their identities and coordination geometries are easy to misassign in structural models. This paper introduces the CheckMyMetal validation service, which detects misidentifications, geometric irregularities, and other implausible features at macromolecular metal-binding sites. The authors organize real examples by metal type to illustrate common problems in crystallographic modeling and offer corresponding modeling and validation recommendations. They note that, at the time, about 40% of macromolecular structures in the PDB contained metal ions, meaning such local errors could affect functional interpretation broadly. CMM turns the chemical plausibility of metal sites into checkable evidence, helping identify positions that need review. Its flags should still be interpreted alongside experimental density, sample conditions, and coordination chemistry; the output alone cannot establish the identity of an unknown ion.
Highlight · CheckMyMetal specifically validates metal identity and coordination geometry, helping uncover site problems that routine checks can miss.
15 · A New Generation of Crystallographic Validation Tools for the Protein Data Bank
Structure · Publication date: 2011
Authors: Randy J. Read, Paul D. Adams, W. Bryan Arendall, Axel T. Brunger, Paul Emsley, Robbie P. Joosten, Gerard J. Kleywegt, Eugene B. Krissinel, Thomas Lütteke, Zbyszek Otwinowski, Anastassis Perrakis, Jane S. Richardson, William H. Sheffler, Janet L. Smith, Ian J. Tickle, Gert Vriend, Peter H. Zwart
The growth of the PDB and mandatory deposition of structure factors created new opportunities to validate models against their underlying diffraction data. This report from the wwPDB X-ray Validation Working Group revisits existing quality standards and stresses that structures can be compared with the database as a whole and with structures at similar resolution, rather than assessed in isolation from a handful of values. The group recommends reporting a small set of easy-to-understand core validation metrics, while retaining details for users who want a deeper analysis, and giving reviewers and editors concise, mature summaries of structural quality. The paper reflects a consensus on validation at the database and peer-review level; it does not propose that any single metric can establish that a structure is entirely correct. Its practical value lies in making quality evidence more transparent and comparable, and helping non-specialists identify structures that warrant closer inspection.
Highlight · Validation reports should provide database-wide and resolution-matched comparisons, alongside a concise set of core quality metrics for reviewers.
16 · MolProbity: More and better reference data for improved all‐atom structure validation
Protein Science · Publication date: 2018
Authors: Christopher J. Williams, Jeffrey J. Headd, Nigel W. Moriarty, Michael G. Prisant, Lizbeth L. Videau, Lindsay N. Deis, Vishal Verma, Daniel A. Keedy, Bradley J. Hintze, Vincent B. Chen, Swati Jain, Steven M. Lewis, W. Bryan Arendall, Jack Snoeyink, Paul D. Adams, Simon C. Lovell, Jane S. Richardson, David C. Richardson
The MolProbity update not only adds checks but also rebuilds the reference data and software infrastructure on which validation depends. This paper describes updated Ramachandran and side-chain rotamer standards, based on roughly a million quality-filtered residues, as well as CaBLAM analysis for checking low-resolution backbones and secondary structure. The program also distinguishes hydrogen distances and van der Waals radii appropriate for X-ray electron-cloud centers versus nuclear positions, and flags rare non-proline cis peptides and twisted peptides. Migration to CCTBX/Python improved maintainability and integration with PHENIX. The authors report that widespread validation and correction have been accompanied by continued improvement in all-atom clashscores for newly submitted structures. These checks primarily assess model plausibility and should still be interpreted alongside the degree of support from experimental data.
Highlight · Improved reference distributions and low-resolution backbone checks help MolProbity pinpoint model outliers in greater detail.
17 · Mg2+ ions: do they bind to nucleobase nitrogens?
Nucleic Acids Research · Publication date: 2017
Authors: Filip Leonarski, Luigi D'Ascenzo, Pascal Auffinger
The way magnesium ions bind in nucleic acid structures is often used to explain catalytic mechanisms, but the chemical identity of a density peak is not always reliable. The authors examined inner-sphere coordination of Mg2+ to the N1, N3 and N7 atoms of bases in the PDB and found that most magnesium assignments near these nitrogens stem from misinterpreted density; the peaks may instead represent sodium, potassium, ammonium, water or artifacts. With a few documented exceptions, Mg2+ generally does not bind directly to purine N7. Manganese, zinc and cadmium, which have a higher affinity for nitrogen, cannot simply be treated as equivalent to magnesium. The article also discusses potential zinc-binding sites and offers recommendations for ion assignment in crystallography, cryo-EM and modeling. The findings are a reminder that metal-substitution experiments and catalytic interpretations must take coordination chemistry into account; ions should not be assigned based on peak position alone.
Highlight · Density peaks near nucleic acid nitrogens should not automatically be assigned as Mg2+; both ion chemistry and the coordination environment need to be checked.
18 · Definition of the hydrogen bond (IUPAC Recommendations 2011)
Pure and Applied Chemistry · Publication date: 2011
Authors: Elangannan Arunan, Gautam R. Desiraju, Roger A. Klein, Joanna Sadlej, Steve Scheiner, Ibon Alkorta, David C. Clary, Robert H. Crabtree, Joseph J. Dannenberg, Pavel Hobza, Henrik G. Kjaergaard, Anthony C. Legon, Benedetta Mennucci, David J. Nesbitt
These IUPAC recommendations propose a new definition of the hydrogen bond, drawing on theoretical and experimental evidence accumulated over the past century. The central point is that identifying a hydrogen bond requires supporting evidence, and the recommendations list six types of criteria that can serve as evidence for its existence. The available abstract is very brief and does not describe the criteria individually, so this issue summarizes only the aim of revising the definition and its evidentiary requirements; it does not fill in details from other sources or general knowledge.
Highlight · Hydrogen bonds should be identified on the basis of supporting evidence; consult the original recommendations for the specific criteria.
Authors: M. Galchenkova, I. Dawod, S. Cardoch, J. Sprenger, D. Oberthür, S. Awel, E. De Santis, O. Grånäs, J. Knoska, M. Metz, S. Passmore, G. Pena, I. Sarrou, M. Wiedorn, N. Timneanu, H. Chapman, O. Yefanov, C. Caleman
Whether femtosecond crystallography can capture reliable structures before radiation damage occurs depends on the pulse conditions, not simply on whether the pulses are short enough. At similar peak intensities and a fixed photon energy, the study compared serial diffraction data from hemoglobin crystals collected with 3- and 10-fs pulses. The refined models and data-quality metrics across resolutions were similar for both datasets, and both retained anomalous signal useful for phasing. Combining collision-radiation and molecular-dynamics simulations, the authors found that under the conditions tested, changes in atomic scattering factors were below 1%, and displacements remained below the resolution limit set by photon energy and detector geometry. Together, the experiments and simulations support the conclusion that, within the intensity range studied, 10-fs pulses are sufficient to obtain protein structural information without significant damage. This finding should not be generalized beyond the specific irradiation conditions.
Highlight · In the hemoglobin experiments, 10-fs pulses preserved structural information comparable to that obtained with shorter pulses under the intensities tested.
Authors: Dorothee Liebschner, Pavel V. Afonine, Nigel W. Moriarty, Billy K. Poon, Oleg V. Sobolev, Thomas C. Terwilliger, Paul D. Adams
Crystallographic maps generally incorporate information from the current model, which can introduce model bias during validation. OMIT maps aim to determine whether atoms are supported by the data by removing them from the model and calculating the residual density. But the omitted region may be filled by a flat bulk-solvent model, further obscuring already weak density. Polder maps keep bulk solvent out of the specified OMIT region, improving the display of weak density for ligands, solvent, side chains, alternate conformations, and residues at termini or in loops. The method has been implemented in PHENIX. Its contribution is to address a specific source of bias in map calculation and improve interpretability—not to make the appearance of density automatic proof of molecular identity. Chemical plausibility and other experimental evidence still need to be considered together.
Highlight · By excluding bulk solvent from the OMIT region, Polder maps help prevent the solvent model from masking weak density.
21 · Linking Crystallographic Model and Data Quality
Science · Publication date: 2012
Authors: P. Andrew Karplus, Kay Diederichs
In crystallography, data quality and model fit are often described using metrics on different scales, making it difficult to judge high-resolution cutoffs and how much room remains for model improvement. The article points out that the widely used Rmerge is not a suitable primary basis for setting the high-resolution limit, and that conventional thresholds can discard data that still have value. The authors introduce CC*, which estimates the correlation between observed data and the underlying true signal that cannot be measured directly, allowing model and data quality to be compared on the same scale. This statistical link helps determine which data are worth retaining and whether further model improvement is already limited by data quality. The work provides a statistically grounded basis for decision-making; it does not advocate keeping all weak data indiscriminately. Cutoff choices still need to be assessed in light of the specific data and refinement results.
Highlight · CC* puts model and data quality on a common scale, helping assess the value of weak data and the limits of model improvement.
After diffraction spots have been integrated, data reduction still involves determining symmetry, scaling observations and assessing data quality. This article describes POINTLESS for symmetry analysis and diagnostic features added to AIMLESS’s scaling model to help determine which measurements are worth retaining. The authors also compare processing statistics, refinement against observed and simulated data, automated model building, and maps at different cutoffs to discuss how to determine the effective resolution. Their tests showed that including weaker high-resolution data beyond commonly used thresholds can bring improvements without causing harm. The key is to judge data by their actual contribution to the model and maps rather than apply a single threshold mechanically. These test results are specific to the data examined and do not mean that data of any quality should always be retained.
Highlight · Resolution cutoffs should be judged by refinement and map quality; weaker data beyond conventional thresholds may still be useful.
Structural resolution is related to coordinate precision, but it cannot be taken directly as a measure of uncertainty in every atomic position. Using full-matrix least squares, the article compares restrained and unrestrained refinement, distinguishes the standard uncertainties of atomic positions from those of bond lengths, and uses high-resolution concanavalin A data to demonstrate how strongly precision varies with B factors. Because inverting the full matrix is computationally expensive, the author also reviews several approximation methods and proposes the diffraction-component precision index (DPI), based on R or Rfree, as a rough tool for comparing structures at different resolutions. The paper also highlights how restraints and disordered regions affect interpretation. Its value lies in framing the question of how precise a model really is in terms of uncertainty, rather than assigning one uniform error value to an entire protein.
Highlight · Precision estimates such as DPI help distinguish structural resolution from coordinate uncertainty and the effects of geometric restraints.
This is an erratum to “Remarks about protein structure precision.” Zotero does not contain an abstract, and no usable abstract was available from the original source. This issue therefore lists only the metadata and does not speculate about the details of the corrections.
Authors: Hesam N. Motlagh, James O. Wrabl, Jing Li, Vincent J. Hilser
Allosteric regulation occurs when a binding event at one site affects a distant functional site. Traditional explanations often focus on a small number of static structures and the transitions between them. Drawing on experimental observations that dynamic and intrinsically disordered proteins can also undergo allostery, this review shifts the focus to conformational ensembles and the statistical properties of interactions. Rather than seek a single structural pathway for transmitting information, it asks how different conformational states and their population distributions jointly encode that information. Analyzing allosteric ensembles reveals diverse regulatory strategies and offers a common framework for comparing different systems. The abstract supports this conceptual perspective and research direction, but does not propose a quantitative model that can be applied directly to every protein; the states, coupling and functional relationships in each system still need to be characterized with experiments and appropriate models.
Highlight · Conformational ensembles and their statistical coupling offer a unified view of allostery in both ordered and disordered proteins.
Many proteins have now been structurally characterized, but a static structural snapshot cannot fully explain how they perform their diverse functions. This article emphasizes the dynamic nature of proteins and argues that structural biology needs a time dimension to describe not only where atoms are, but also when they move. The authors frame the goal as observing protein activity in real time at atomic resolution, bringing structural descriptions closer to the processes that underlie function. The abstract primarily sets out this vision and does not specify experimental systems, time resolution or performance results, so we do not present it as evidence that a particular new technology has already achieved the full goal. It offers a conceptual entry point for understanding the relationship between conformational motion and function, and pairs well with this issue’s papers on conformational ensembles, room-temperature structures and ligand-binding pathways.
Highlight · Understanding protein function requires adding a time dimension to static three-dimensional structures to describe motion.
27 · Conformational selection or induced fit: A flux description of reaction mechanism
Proceedings of the National Academy of Sciences · Publication date: 2009
Authors: Gordon G. Hammes, Yu-Chu Chang, Terrence G. Oas
When ligand binding involves a conformational change, simply comparing single-step rate constants cannot establish whether the conformation changes first or the ligand binds first. This article proposes measuring the relative contributions of conformational selection and induced fit by the reaction flux through each pathway, and provides rules for flux analysis of multistep mechanisms. The authors reanalyze two cases: NADPH binding to dihydrofolate reductase, and flavin mononucleotide binding alongside flavodoxin folding. They find that conformational selection predominates at low ligand concentrations, while induced fit predominates at high concentrations; across a broad range of conditions, both pathways carry substantial flux. Thus, a mechanism may not fit neatly into one of two fixed categories, and the balance can shift with experimental conditions. This method offers a quantitative way to interpret kinetic data, but the contributions of particular pathways must still be calculated from data for the system in question.
Highlight · Comparing pathway fluxes, rather than individual rate constants, is key to assessing the relative contributions of conformational selection and induced fit.
28 · On the nature of allosteric transitions: A plausible model
Journal of Molecular Biology · Publication date: 1965
Authors: Jacques Monod, Jeffries Wyman, Jean-Pierre Changeux
Zotero does not contain an abstract, and attempts to retrieve one from the DOI landing page and PubMed were unsuccessful. This issue lists only the metadata and does not infer the model’s contents from the title.
29 · Accessing protein conformational ensembles using room-temperature X-ray crystallography
Proceedings of the National Academy of Sciences · Publication date: 2011
Authors: James S. Fraser, Henry van den Bedem, Avi J. Samelson, P. Therese Lang, James M. Holton, Nathaniel Echols, Tom Alber
The fact that the overall protein backbone changes little after crystal cooling does not mean that functionally relevant conformational ensembles are fully preserved. Using electron-density sampling, model refinement, and packing-analysis tools, the study compared data from 30 proteins and found that flash cooling can reshape the conformational distributions of more than 35% of side chains and eliminate packing defects associated with functional motions. Analysis of H-Ras also revealed an allosteric network in room-temperature electron density that matched fluctuations observed by solution NMR, but was not apparent in cryogenic maps. These findings suggest that structural databases may be biased toward tighter, more densely packed, apparently single-conformation models. Room-temperature crystallography can therefore add dynamic information relevant to catalysis, ligand binding, and allosteric regulation. The point is that cooling can bias the observed ensemble—not that cryogenic structures have no value.
Highlight · Room-temperature crystallography can reveal conformational ensembles and functional motions that cooling alters or obscures.
AlphaFold Structure Prediction and Evaluation of Its Applications (5 papers)
30 · Accurate structure prediction of biomolecular interactions with AlphaFold 3
Nature · Publication date: 2024
Authors: Josh Abramson, Jonas Adler, Jack Dunger, Richard Evans, Tim Green, Alexander Pritzel, Olaf Ronneberger, Lindsay Willmore, Andrew J. Ballard, Joshua Bambrick, Sebastian W. Bodenstein, David A. Evans, Chia-Chun Hung, Michael O’Neill, David Reiman, Kathryn Tunyasuvunakool, Zachary Wu, Akvilė Žemgulytė, Eirini Arvaniti, Charles Beattie, Ottavia Bertolli, Alex Bridgland, Alexey Cherepanov, Miles Congreve, Alexander I. Cowen-Rivers, Andrew Cowie, Michael Figurnov, Fabian B. Fuchs, Hannah Gladman, Rishub Jain, Yousuf A. Khan, Caroline M. R. Low, Kuba Perlin, Anna Potapenko, Pascal Savy, Sukhdeep Singh, Adrian Stecula, Ashok Thillaisundaram, Catherine Tong, Sergei Yakneen, Ellen D. Zhong, Michal Zielinski, Augustin Žídek, Victor Bapst, Pushmeet Kohli, Max Jaderberg, Demis Hassabis, John M. Jumper
AlphaFold 3 expands the scope of structure prediction beyond protein-centered systems to a much broader range of biomolecular complexes. Its substantially updated diffusion architecture predicts structures jointly for proteins, nucleic acids, small molecules, ions, and modified residues. On the benchmarks used in the paper, the authors report better protein–ligand predictions than leading docking tools, better protein–nucleic acid interaction predictions than specialized predictors, and better antibody–antigen predictions than AlphaFold-Multimer v2.3. A major advance is bringing several interaction types that were previously handled separately into a unified deep-learning framework. These comparisons apply to the specific datasets and tasks tested; they do not mean that every complex system can be predicted reliably. Particular binding modes still need to be assessed alongside confidence scores and independent experimental evidence.
Highlight · Its diffusion architecture lets AlphaFold 3 predict structures involving diverse biomolecules in a unified framework.
31 · Addendum: Accurate structure prediction of biomolecular interactions with AlphaFold 3
Nature · Publication date: 2024
Authors: Josh Abramson, Jonas Adler, Jack Dunger, Richard Evans, Tim Green, Alexander Pritzel, Olaf Ronneberger, Lindsay Willmore, Andrew J. Ballard, Joshua Bambrick, Sebastian W. Bodenstein, David A. Evans, Chia-Chun Hung, Michael O’Neill, David Reiman, Kathryn Tunyasuvakool, Zachary Wu, Akvilė Žemgulytė, Eirini Arvaniti, Charles Beattie, Ottavia Bertolli, Alex Bridgland, Alexey Cherepanov, Miles Congreve, Alexander I. Cowen-Rivers, Andrew Cowie, Michael Figurnov, Fabian B. Fuchs, Hannah Gladman, Rishub Jain, Yousuf A. Khan, Caroline M. R. Low, Kuba Perlin, Anna Potapenko, Pascal Savy, Sukhdeep Singh, Adrian Stecula, Ashok Thillaisundaram, Catherine Tong, Sergei Yakneen, Ellen D. Zhong, Michal Zielinski, Augustin Žídek, Victor Bapst, Pushmeet Kohli, Max Jaderberg, Demis Hassabis, John M. Jumper
This is an addendum to the original AlphaFold 3 paper. Zotero has no abstract saved, and the original publication page provides no usable abstract. This issue therefore lists the metadata only and does not speculate about the addendum’s contents.
32 · AlphaFold predictions are valuable hypotheses and accelerate but do not replace experimental structure determination
Nature Methods · Publication date: 2024
Authors: Thomas C. Terwilliger, Dorothee Liebschner, Tristan I. Croll, Christopher J. Williams, Airlie J. McCoy, Billy K. Poon, Pavel V. Afonine, Robert D. Oeffner, Jane S. Richardson, Randy J. Read, Paul D. Adams
Whether high confidence in an AlphaFold model guarantees that its local details are correct is a key question when using predicted structures. The authors directly compared predicted models with experimental crystallographic maps and found that many predictions agree closely with experiment, but some high-confidence models still show overall distortions, incorrect domain orientations, or mismatches in local backbone and side-chain conformations. They also note that the predictions examined did not account for ligands, covalent modifications, or other environmental factors. The authors recommend treating predictions as highly valuable structural hypotheses and interpreting them in light of confidence scores, with experimental validation especially important for interactions and key details not included in the prediction. The study does not discount the role of predictions in accelerating structure determination; rather, it clarifies the boundary between a useful hypothesis and experimentally verified detail.
Highlight · Even high-confidence predictions can miss experimental details; AlphaFold models are best treated as structural hypotheses to be validated.
33 · Highly accurate protein structure prediction with AlphaFold
Nature · Publication date: 2021
Authors: John Jumper, Richard Evans, Alexander Pritzel, Tim Green, Michael Figurnov, Olaf Ronneberger, Kathryn Tunyasuvakool, Russ Bates, Augustin Žídek, Anna Potapenko, Alex Bridgland, Clemens Meyer, Simon A. A. Kohl, Andrew J. Ballard, Andrew Cowie, Bernardino Romera-Paredes, Stanislav Nikolov, Rishub Jain, Jonas Adler, Trevor Back, Stig Petersen, David Reiman, Ellen Clancy, Michal Zielinski, Martin Steinegger, Michalina Pacholska, Tamas Berghammer, Sebastian Bodenstein, David Silver, Oriol Vinyals, Andrew W. Senior, Koray Kavukcuoglu, Pushmeet Kohli, Demis Hassabis
The number of known protein sequences far exceeds the number of experimentally determined structures, making reliable sequence-to-structure prediction fundamental. This paper introduces a redesigned AlphaFold neural network. In the CASP14 blind assessment, it achieved accuracy competitive with experimental structures for most targets, including high accuracy when no similar known structure was available. The method incorporates physical and biological knowledge of protein structure, along with information from multiple sequence alignments, into a deep-learning architecture, reducing reliance on existing homologous templates. This work greatly broadened access to structural research and laid the groundwork for large-scale structural bioinformatics. The abstract reports overall performance in a blind assessment; it does not mean that every residue, state, or arbitrary complex reaches experimental accuracy. The uncertainty of each prediction still needs to be evaluated.
Highlight · AlphaFold demonstrated at CASP14 that it can predict highly accurate structures from sequence, even without relying on homologous templates.
Authors: Mehmet Akdel, Douglas E. V. Pires, Eduard Porta Pardo, Jürgen Jänes, Arthur O. Zalevsky, Bálint Mészáros, Patrick Bryant, Lydia L. Good, Roman A. Laskowski, Gabriele Pozzati, Aditi Shenoy, Wensi Zhu, Petras Kundrotas, Victoria Ruiz Serra, Carlos H. M. Rodrigues, Alistair S. Dunham, David Burke, Neera Borkakoti, Sameer Velankar, Adam Frost, Jérôme Basquin, Kresten Lindorff-Larsen, Alex Bateman, Andrey V. Kajava, Alfonso Valencia, Sergey Ovchinnikov, Janani Durairaj, David B. Ascher, Janet M. Thornton, Norman E. Davey, Amelie Stein, Arne Elofsson, Tristan I. Croll, Pedro Beltrao
A prediction that closely resembles an experimental structure is not automatically suitable for every structural biology task. This community assessment examines AlphaFold2’s applications in identifying structural features, interpreting missense variants, predicting function and ligand-binding sites, modeling interactions, and building models from experimental data. Analysis of 11 proteomes found that, on average, about 25% more residues could be modeled with confidence than with comparative modeling, and revealed structural features rarely seen in the PDB. The study also assesses disordered regions and complex predictions, concluding that AF2 models can be as useful as experimental models for many tasks, provided confidence metrics are interpreted critically. The focus is on evaluating real-world applications and their limitations—not on treating all predictions as substitutes for experimental structures.
Highlight · AF2 expands the range of structures that can be modeled with confidence, but its confidence metrics still need careful interpretation across tasks.
35 · Molecular Pharmacology of Dopamine Receptors: Structure, Function, and Therapeutic Implications
Annual Review of Pharmacology and Toxicology · Publication date: 2026
Authors: Youwen Zhuang, H. Eric Xu
Dopamine receptors have long been targets for drug development, but available treatments remain limited by inadequate subtype selectivity, insufficient specificity in pathway engagement, and adverse effects. Using active and inactive receptor structures as a framework, this review brings together structural, biochemical, pharmacological, and computational studies to discuss activation, inverse agonism, transducer coupling, biased signaling, allosteric modulation, and ligand selectivity. The authors focus in particular on how ligand-induced conformations, transducer interactions, and the membrane environment jointly shape signaling outputs, and compare activation pathways in D1-like and D2-like receptors. The review also considers what allosteric sites and AI-assisted discovery may mean for drug design. Its aim is to move from descriptive pharmacology toward a mechanistic, structure-guided design framework—not to claim that a new therapy has already improved clinical safety or efficacy.
Highlight · Drug selectivity at dopamine receptors must be understood in terms of conformation, coupling partners, and the membrane environment.
Authors: Richa Agrawal, Ramon Mendoza Uriarte, Bernardo I. Pinto-Anwandter, Trayder Thomas, Lydia Blachowicz, Young Hoon Koh, Francisco Bezanilla, Eduardo Perozo, Benoît Roux
How voltage-gated potassium channels convert movement of the voltage sensor into pore opening is a central question in electromechanical coupling. The authors determined the cryo-EM structure of the Shaker channel ILT mutant, which partially uncouples gating-charge movement in the voltage-sensing domain from pore opening. This captured an intermediate state in which the sensor is partly activated while the pore gate remains closed. Combined with AlphaFold2-based modeling and molecular dynamics simulations, the study proposes that shifts in the population balance between two metastable positions of the S4–S5 linker transmit sensor activation to the pore. This result brings dynamic conformational distributions into the coupling mechanism, rather than relying solely on comparisons between two endpoint structures. The evidence comes from a specific mutant system and computational analyses; its generality to other channels requires further validation.
Highlight · Shifts between metastable states of the S4–S5 linker offer a dynamic explanation for coupling between the voltage sensor and pore.
37 · GPCR antagonism via rewiring of receptor trafficking and degradation
Nature · Published/issue date: 2026-09-30
Authors: Kaitlin Rhee, Lawrence Shue, Akimasa Adachi, Pengwei Sun, James Osei-Owusu, Dingjingyu Zhou, Aoxing Cheng, Yi Ran Xu, Qingyue Li, Apoorva Baluapuri, Edward P. Harvey, Karen Adelman, Meredith A. Skiba, Bianxiao Cui, Jun R. Huh, Andrew C. Kruse, Xin Zhou
Conventional GPCR antagonists mainly block receptor activity; this study instead redirects receptor trafficking and degradation. The authors engineered a bispecific antibody chimera, GTAC, that recognizes both GPCRs and the transferrin receptor TfR1, prompting target-receptor internalization and routing it to lysosomal degradation. Across tested targets including BILF1, RXFP1, and CCR6, GTAC selectively reduced receptor levels and suppressed receptor output, including constitutive signaling. Its reported potency was more than one to two orders of magnitude greater than that of conventional antibody antagonists. Protein engineering and multicolor live-cell imaging further clarified the cellular mechanism and context-dependent design requirements. The work shows that altering receptor fate can provide a functional antagonism strategy, but the reported potency comes from experimental systems and cannot be equated directly with therapeutic benefit in humans.
Highlight · Routing GPCRs to internalization and lysosomal degradation enables functional antagonism by reducing receptor abundance.
Authors: Kevin M. Jude, Carl-Mikael Suomivuori, Deepa Waghray, Shoji Maeda, Yoshinori Fujiyoshi, Asuka Inoue, K. Christopher Garcia, Naotaka Tsutsumi
The process by which GPCRs recognize G proteins and trigger GDP release is difficult to reconstruct from a single stable complex structure. Using the viral chemokine receptor US28, the researchers obtained cryo-EM structures of three US28–Gq states: a GDP-bound encounter complex, the canonical nucleotide-free state, and a possible intermediate linking the two. Together with simulations and functional data, these structures support a conformational model for stepwise G-protein activation and echo previously proposed mechanisms for human GPCR activation. The work provides several high-resolution snapshots that support a possible trajectory; it does not directly observe the entire activation process continuously. Its main significance is to connect structural changes from initial recognition to subsequent nucleotide release, providing a basis for comparing receptor signaling mechanisms.
Highlight · Multiple US28–Gq structures outline a stepwise model of G-protein activation, from initial contact to nucleotide release.
Brain activity and mechanisms of neurological disorders (3 papers)
39 · Psychedelics align brain activity with context
Nature · Published/issue date: 2026-08
Authors: Devon Stoliker, Leonardo Novelli, Moein Khajehnejad, Mana Biabani, Matthew D. Greaves, Tamrin Barta, Martin Williams, Sidhant Chopra, Olivier Bazin, Otto Simonsson, Richard Chambers, Frederick S. Barrett, Gustavo Deco, Katrin H. Preller, Robin L. Carhart-Harris, Anil K. Seth, Suresh Sundram, Gary F. Egan, Adeel Razi
Brain activity under psychedelics is often described as more disordered, but time-averaged measures may miss its underlying organization. The study recorded functional MRI and EEG from 62 adults, comparing activity before dosing with activity on the day of psilocybin administration during rest, meditation, music, and film. Using machine learning to represent each person's brain dynamics as low-dimensional trajectories, the authors found that activity under the drug still showed structured patterns aligned with context; networks typically associated with internal and external processing were more integrated. This degree of contextual alignment was associated with ego dissolution, feelings of merging with the environment, and changes in mood the following day. The findings offer a dynamic account of apparent “disorder,” but correlations between experience and brain activity do not establish long-term therapeutic effects or a single causal mechanism.
Highlight · Low-dimensional trajectories reveal a hidden order in brain activity under psilocybin, structured by context.
40 · Circuit mechanisms of psychotic symptoms across disorders
Neuron · Published/issue date: 2026/10/01
Authors: Urs Braun, Johannes Wolf, Dusan Hirjak, Emanuel Schwarz, Heike Tost, Andreas Meyer-Lindenberg
Hallucinations, delusions, and disorganized thinking occur across a range of psychiatric and neurological disorders, yet there is no unified explanation of how molecular abnormalities give rise to symptoms. This article proposes a transdiagnostic framework: disruptions in dopamine, glutamate, GABA, serotonin, and acetylcholine alter neuronal gain and synaptic efficacy, affecting the precision weighting of sensory evidence, prediction errors, and internal representations. These local changes propagate through cortico-striato-thalamo-cortical circuits, leading to abnormal integration and segregation in large-scale networks and unstable attractor states. The model allows different biological pathways to converge on shared computational vulnerabilities and proposes individualized testing using imaging, dynamical-systems methods, and AI. This is a mechanistic synthesis and a testable framework, not a validated diagnostic or treatment tool.
Highlight · A cross-scale framework links neurotransmitter disruptions and circuit dynamics to psychotic symptoms shared across disorders.
41 · Prion-like transmission of human tau strains in the mouse brain
Nature · Published/issue date: 2026-09-30
Authors: Sofia Lövestam, Aki Shimozawa, Airi Tarutani, Reiko Ohtani, Masami Masuda-Suzukake, Kazuko Hasegawa, Andrew C. Robinson, Yuko Saito, Shigeo Murayama, Mari Yoshida, Hisaomi Suzuki, Mitsumoto Onaya, Masato Hasegawa, Michel Goedert, Sjors H. W. Scheres
Neurodegeneration-associated fibrils are thought to spread through templated seeding, but direct evidence is needed to determine whether distinct “strains” retain their original structures after transmission. The researchers injected tau fibrils from the brains of people with Alzheimer’s disease or corticobasal degeneration into the brains of wild-type mice. Newly formed amyloid fibrils made of mouse tau were found to share the same structures as the human-derived seeds. The results support the idea that tau folds can be replicated by templating and provide structural evidence that distinct conformations correspond to different pathological types. The authors propose that mice can be used to study the molecular mechanisms by which specific tau folds drive pathology. Here, “prion-like” describes templated conformational transmission in an experiment; it does not imply that these neurodegenerative diseases are contagious through human contact.
Highlight · Human tau seeds induced the same fold in mice, supporting templated transmission of strain-specific structural identity.
42 · A normalized root‐mean‐spuare distance for comparing protein three‐dimensional structures
Protein Science · Published/issue date: 2001
Authors: Oliviero Carugo, Sándor Pongor
Root-mean-square deviation (RMSD) is a widely used measure for comparing protein structures, but its value also depends on the number of equivalent atom pairs included in the alignment. As a result, two structure pairs of markedly different lengths may have the same RMSD without being equally similar. This paper proposes a simple normalization procedure intended to reduce this size dependence and make comparisons across structures of different sizes more comparable. The authors suggest that the approach could be used in studies of structural evolution, fold classification, and direct comparison of structural models. The abstract does not provide the formula or range of applicability, so we do not add operational details here. The paper serves as a reminder that when citing a single measure of structural difference, the alignment size and normalization method should also be specified rather than ranking all RMSD values on the same scale.
Highlight · Structural comparisons need to account for alignment size; normalized RMSD aims to reduce size-related bias.
43 · TM-align: a protein structure alignment algorithm based on the TM-score
Nucleic Acids Research · Published/issue date: 2005
Authors: Yang Zhang, Jeffrey Skolnick
TM-align combines a TM-score rotation matrix with dynamic programming to find structural alignments between two protein chains. In the paper’s benchmarks, it was faster than CE, DALI, and SAL, and performed better in average alignment accuracy and coverage. The authors also conducted pairwise comparisons of more than 10,000 representative PDB chains, classified structural folds at a specified TM-score threshold, and tested the similarity of predicted models to known structures. One key observation was that both correctly and incorrectly folded models could have close structural analogs, so finding a similar structure alone is not proof that a prediction is correct. However, similarity still correlates with model accuracy in ways that can help with model selection. The tool’s value lies in efficiently quantifying structural similarity, while interpretation depends on the alignment range and intended use.
Highlight · TM-align efficiently measures structural similarity, while reminding us that finding a similar structure does not prove a model is correct.
44 · Inference of Macromolecular Assemblies from Crystalline State
Journal of Molecular Biology · Published/issue date: 2007
Authors: Evgeny Krissinel, Kim Henrick
Symmetry contacts in crystal structures may reflect genuine biological assemblies or merely crystal packing, and distinguishing between the two matters for functional interpretation. This paper discusses the physicochemical principles underlying stable macromolecular complexes and methods for calculating affinity and assembly entropy, the latter of which has a major effect on complex size and symmetry. On this basis, the authors developed a method grounded in chemical thermodynamics to automatically identify likely assemblies from PDB entries derived from X-ray diffraction, and implemented it as a public web service. The paper reports a success rate of about 80–90% for recovering biological units in its evaluation. This provides a tool for studying protein interactions using crystal structures, while also underscoring the uncertainty in such inferences: particular assemblies, especially those involving weak interactions, should still be assessed alongside other evidence.
Highlight · Thermodynamics-based assembly inference helps distinguish biological complexes from crystal packing, but predictions are not certain in every case.
45 · Macromolecular complexes in crystals and solutions
Acta Crystallographica Section D Biological Crystallography · Published/issue date: 2011
Author: Evgeny Krissinel
Contacts in a crystal do not always faithfully reflect biological assemblies in solution. This paper reviews established approaches to analyzing macromolecular interactions and complexes from crystal packing, and identifies conditions under which routine inference can fail. Computational analyses indicate that weak interactions with dissociation constants of around 100 μM or higher are likely to be lost during crystallization. This means the risk of disagreement between crystal structures and native assemblies cannot be ignored for some PDB dimers. The author estimates that for about one-fifth of protein dimers, the probability of such a discrepancy exceeds 50%, and recommends supplementing crystal-based inference with non-crystallographic studies when assessing weakly bound complexes. These percentages are computational estimates from the paper; they are best understood as a caution about methodological limits, not as a direct verdict that any particular database assembly is wrong.
Highlight · Weakly bound complexes may change during crystallization, so crystal-based assembly predictions should be supported by solution studies or other independent evidence.
46 · Bluues2: An enhanced web server for fast and accurate electrostatic analysis of proteins and protein complexes
Protein Science · Published/issue date: 2026
Authors: Damiano Piovesan, Nima Rasekh, Miguel Angel Soler, Walter Rocchia, Gennaro Esposito, Federico Fogolari, Silvio C. E. Tosatto
Electrostatic interactions affect protein stability, binding, and function, but analyzing complexes often means switching between computational and visualization tools. Bluues2 updates a service for protein electrostatics based on the generalized Born model. It uses NanoShaper to generate molecular surfaces, improves GB radius calculations, and extends analysis to protein–protein and protein–ligand complexes. The service calculates pKa, surface potential, solvation free energy, binding free energy, and interfacial electrostatic complementarity, while integrating the Mol* viewer and DRMAAtic task management. Its contribution is to bring these analyses together in a single interface, making it easier to interpret structures and interface properties. The abstract provides no quantitative benchmarks for improved accuracy, however, so it does not establish that the service outperforms other methods across all systems.
Highlight · Bluues2 brings a range of electrostatic analyses for proteins and complexes together in a single visualization service.
47 · The gasdermin family: from pyroptosis mechanisms to therapeutic targets
Signal Transduction and Targeted Therapy · Published: October 2, 2026
Authors: Weilv Xu, Shiyang Liu, Zexu Yu, Jinhuang Shi, Yang Yang, Fushan Shi
Gasdermins are key executors of pyroptosis, but their roles extend beyond inflammatory membrane rupture. This review surveys how family members are activated and regulated, how they form pores and drive cell rupture, and how membranes are repaired. It also covers functions that may or may not depend on pyroptosis, including mucus secretion, bone resorption, and mitochondrial metabolism. The authors discuss crosstalk with apoptosis, necroptosis, NETosis, and PANoptosis, as well as context-dependent roles in host defense, chronic inflammation, and tumor immunity. The therapeutic section reviews small molecules such as disulfiram and material-based delivery strategies. The review emphasizes that this family can both protect the host and drive disease, so interventions must account for tissue and disease context. The strategies discussed reflect research progress and potential, not therapies that have all been clinically validated.
Highlight · Understanding the therapeutic potential of gasdermins requires accounting for functions beyond pyroptosis and their context-dependent effects.
48 · Function-preserving watermarking of AI-generated proteins
Nature · Published: September 30, 2026
Authors: David Stutz, Alexander I. Cowen-Rivers, Guillermo Ortiz-Jimenez, Jeremy Ratcliff, Vinicius Zambaldi, Lindsay Willmore, Josh Abramson, Harshnira Patani, Christina Kouridi, Florian Stimberg, Mel Vecerik, Alex Chu, Sukhdeep Singh, Sumanth Dathathri, Eliseo Papa, Valentin De Bortoli, Arnaud Doucet, Demis Hassabis, Jue Wang, Sven Gowal, Pushmeet Kohli
Generative AI is accelerating protein design, but it also makes it harder to trace the origins of sequences and structures. The authors introduce SynthIDBio, which embeds detectable watermarks in both protein sequences and predicted structures. The sequence approach adds an identifier while preserving function: experimentally designed binding proteins showed binding affinities similar to those of unwatermarked controls, and watermark detection accuracy was nearly perfect in tests. The structural approach fine-tunes AlphaFold3 to embed subtle watermarks in biomolecular structures. By treating traceability and function preservation as joint design goals, the study provides a proof of concept for tracking the provenance of AI-engineered biological products. The findings do not establish that every protein or downstream transformation has been tested, nor that watermarks cannot be removed by arbitrary modifications.
Highlight · SynthIDBio demonstrates the feasibility of watermarking AI-generated protein sequences and structures without compromising function.
49 · Soft drugs: design principles and topical applications
Nature Reviews Drug Discovery · Published: September 30, 2026
Authors: Tracey Pirali, Alessandro Accetta, Jarkko Rautio, Laura Carzaniga
Soft-drug design aims for a drug to exert its effects first, then become inactive rapidly through a predictable metabolic pathway, helping control duration of action and systemic exposure. This review revisits the design principles and distinguishes soft drugs from prodrugs, which require metabolic activation. For systemic use, rapidly waning effects can be useful in settings such as anesthesia; for topical use, the goal is to retain efficacy at the site of administration while limiting toxicity after systemic absorption. The authors review clinical progress in dermatological, inhaled, and gut-restricted therapies, and discuss practical ways to incorporate soft-drug principles into drug discovery. They argue that the strategy remains underused and is often confused with other concepts. Its value lies in making metabolic inactivation an intentional design goal—not in assuming that faster metabolism is always better. Safety and efficacy still depend on the drug and how it is administered.
Highlight · Soft drugs are designed to become predictably inactive after acting, balancing local efficacy, duration of action, and systemic exposure.
50 · Uniform thin ice on ultraflat graphene for high-resolution cryo-EM
Nature Methods · Published: January 2023
Authors: Liming Zheng, Nan Liu, Xiaoyin Gao, Wenqing Zhu, Kun Liu, Cang Wu, Rui Yan, Jincan Zhang, Xin Gao, Yating Yao, Bing Deng, Jie Xu, Ye Lu, Zhongmin Liu, Mengsen Li, Xiaoding Wei, Hong-Wei Wang, Hailin Peng
Uneven ice thickness during cryo-EM sample preparation can compromise image quality, particularly limiting high-resolution analysis of small-molecular-weight samples. The study found that uniform thin ice is linked to the flatness of the support film and used ultraflat graphene as a support to improve control of vitreous ice thickness. Using this method, the authors determined structures of 64-kDa hemoglobin, asymmetric 67-kDa alpha-fetoprotein, and 52-kDa streptavidin at resolutions of 3.5, 2.6, and 2.2 Å, respectively. The results show that support flatness can be a key variable in optimizing sample preparation. The paper also points to potential applications in cryo-electron tomography and structure-guided drug discovery, but these prospects should be distinguished from the demonstrated single-particle results for small proteins; not all samples can be assumed to benefit equally.
Highlight · Ultraflat graphene improves thin-ice uniformity, helping produce high-quality cryo-EM structures of several low-molecular-weight samples.
51 · In-cell discovery and characterization of a non-canonical bacterial protein translocation-folding complex
Cell · Published: September 25, 2026
Authors: Rasmus K. Jensen, Liang Xue, Federico Marotta, Joseph C. Somody, Joel Selkrig, Swantje Lenz, Juri Rappsilber, Mikhail M. Savitski, Jan Kosinski, Athanasios Typas, Maria Zimmermann-Kogadeeva, Peer Bork, Julia Mahamid
In-cell structural imaging can reveal previously unknown complexes, but establishing their composition and function requires multiple lines of evidence. Using genome-reduced Mycoplasma pneumoniae as a model, the researchers mapped an unknown complex on the cell surface in situ by cryo-electron tomography. They combined this with proteomics, structure prediction, integrative modeling, and bioinformatics to identify a conserved Sec translocation system paired with an extracellular dome-like structure. The latter consists mainly of three previously uncharacterized Mdp proteins and is structurally homologous to the ATP-independent foldase PrsA; Mdp444 retains key catalytic residues and activity. The subnanometer-resolution map links SecA-mediated cotranslational translocation to an extracellular protein-folding system, demonstrating how in situ structural analysis combined with molecular identification can help explain unknown cellular machinery.
Highlight · In situ tomography, combined with omics and prediction, links the bacterial Sec translocation machinery to an extracellular protein-folding system.
This issue features 51 papers spanning cryo-EM reconstruction and validation, atomic models and chemical validation, crystallographic data quality, AlphaFold structure prediction, protein dynamics and allostery, membrane receptors and ion channels, neuroscience, structural alignment, and biological assemblies.
Cryo-EM reconstruction, resolution, and validation (11 papers)
01 · RELION: Implementation of a Bayesian approach to cryo-EM structure determination
Journal of Structural Biology · Publication / issue date: 2012
Authors: Sjors H.W. Scheres
Single-particle cryo-EM reconstruction often involves many parameters that must be tuned by experience. RELION frames this problem in a Bayesian statistical framework. The software estimates statistical model parameters from the data and refines macromolecular structures using maximum a posteriori probability, reducing the need for repeated manual tuning. The paper describes an implementation that lowers computational costs and examines how precisely the relative orientations of individual particles can be determined. It also incorporates gold-standard Fourier shell correlation to guard against overfitting. The authors report that the implementation can produce high-quality reconstructions and reliable resolution estimates at acceptable computational cost, with limited user intervention. Its key contribution is to place parameter estimation, reconstruction, and validation within explicit statistical assumptions. In practice, results still depend on the input data and optimization conditions, so automation should not be mistaken for a substitute for inspection.
Highlight · RELION combines Bayesian parameter estimation with independent half-data validation, enabling single-particle reconstruction with less manual tuning.
DOI: 10.1016/j.jsb.2012.09.006
02 · New tools for the analysis and validation of cryo-EM maps and atomic models
Acta Crystallographica Section D Structural Biology · Publication / issue date: 2018
Authors: Pavel V. Afonine, Bruno P. Klaholz, Nigel W. Moriarty, Billy K. Poon, Oleg V. Sobolev, Thomas C. Terwilliger, Paul D. Adams, Alexandre Urzhumtsev
As the number of atomic models derived from cryo-EM has grown rapidly, simply reusing crystallographic checks has proved insufficient to capture the specific characteristics of maps. This paper presents tools in PHENIX for cryo-EM analysis and validation, including d99 for estimating resolution, phenix.auto_sharpen for improving map appearance, and phenix.mtriage for analyzing map properties. It also demonstrates their use on structures and maps deposited in public archives. The authors recommend retaining half-maps and masks so that outside users can more fully assess reconstructions and atomic models. The work connects map processing and resolution description with model validation, and underscores the importance of making data available for review. The abstract does not prescribe a fixed set of processing parameters for every sample, so these tools should be treated as complementary diagnostics and used in light of the original data and local features.
Highlight · Cryo-EM map tools, together with archived half-maps and masks, support structural validation that others can independently review.
DOI: 10.1107/s2059798318009324
03 · Outcomes of the EMDataResource cryo-EM Ligand Modeling Challenge
Nature Methods · Publication / issue date: 2024
Authors: Catherine L. Lawson, Andriy Kryshtafovych, Grigore D. Pintilie, Stephen K. Burley, Jiří Černý, Vincent B. Chen, Paul Emsley, Alberto Gobbi, Andrzej Joachimiak, Sigrid Noreng, Michael G. Prisant, Randy J. Read, Jane S. Richardson, Alexis L. Rohou, Bohdan Schneider, Benjamin D. Sellers, Chenghua Shao, Elizabeth Sourial, Chris I. Williams, Christopher J. Williams, Ying Yang, Venkat Abbaraju, Pavel V. Afonine, Matthew L. Baker, Paul S. Bond, Tom L. Blundell, Tom Burnley, Arthur Campbell, Renzhi Cao, Jianlin Cheng, Grzegorz Chojnowski, K. D. Cowtan, Frank DiMaio, Reza Esmaeeli, Nabin Giri, Helmut Grubmüller, Soon Wen Hoh, Jie Hou, Corey F. Hryc, Carola Hunte, Maxim Igaev, Agnel P. Joseph, Wei-Chun Kao, Daisuke Kihara, Dilip Kumar, Lijun Lang, Sean Lin, Sai R. Maddhuri Venkata Subramaniya, Sumit Mittal, Arup Mondal, Nigel W. Moriarty, Andrew Muenks, Garib N. Murshudov, Robert A. Nicholls, Mateusz Olek, Colin M. Palmer, Alberto Perez, Emmi Pohjolainen, Karunakar R. Pothula, Christopher N. Rowley, Daipayan Sarkar, Luisa U. Schäfer, Christopher J. Schlicksup, Gunnar F. Schröder, Mrinal Shekhar, Dong Si, Abhishek Singharoy, Oleg V. Sobolev, Genki Terashi, Andrea C. Vaiana, Sundeep C. Vedithi, Jacob Verburgt, Xiao Wang, Rangana Warshamanage, Martyn D. Winn, Simone Weyand, Keitaro Yamashita, Minglei Zhao, Michael F. Schmid, Helen M. Berman, Wah Chiu
A near-atomic-resolution map may reveal a ligand, but that does not mean different modelers will arrive at equally reliable binding poses. The EMDataResource ligand-modeling challenge used three maps at 1.9–2.5 Å resolution, covering inhibitor, covalent nucleotide analog, and lipid-binding cases. Seventeen independent teams submitted 61 models and workflows. The study assessed ligand and surrounding-atom modeling using visual inspection, local map quality, map fit, geometry, energy, and contact scores. The results show that no single score can capture the overall quality of a macromolecule and its ligand; multiple metrics need to be considered together. The paper therefore proposes practices for evaluating ligand structures at near-atomic resolution, emphasizing that neither global resolution nor a good fit score alone guarantees a correct local chemical interpretation.
Highlight · Ligand models should be evaluated using multiple lines of evidence, including map fit, geometry, energy, and contacts.
DOI: 10.1038/s41592-024-02321-7
04 · Cryo-EM model validation recommendations based on outcomes of the 2019 EMDataResource challenge
Nature Methods · Publication / issue date: 2021
Authors: Catherine L. Lawson, Andriy Kryshtafovych, Paul D. Adams, Pavel V. Afonine, Matthew L. Baker, Benjamin A. Barad, Paul Bond, Tom Burnley, Renzhi Cao, Jianlin Cheng, Grzegorz Chojnowski, Kevin Cowtan, Ken A. Dill, Frank DiMaio, Daniel P. Farrell, James S. Fraser, Mark A. Herzik, Soon Wen Hoh, Jie Hou, Li-Wei Hung, Maxim Igaev, Agnel P. Joseph, Daisuke Kihara, Dilip Kumar, Sumit Mittal, Bohdan Monastyrskyy, Mateusz Olek, Colin M. Palmer, Ardan Patwardhan, Alberto Perez, Jonas Pfab, Grigore D. Pintilie, Jane S. Richardson, Peter B. Rosenthal, Daipayan Sarkar, Luisa U. Schäfer, Michael F. Schmid, Gunnar F. Schröder, Mrinal Shekhar, Dong Si, Abishek Singharoy, Genki Terashi, Thomas C. Terwilliger, Andrea Vaiana, Liguo Wang, Zhe Wang, Stephanie A. Wankowicz, Christopher J. Williams, Martyn Winn, Tianqi Wu, Xiaodi Yu, Kaiming Zhang, Helen M. Berman, Wah Chiu
The 2019 cryo-EM model challenge set out to assess the quality of models produced by existing software, how reproducible results were across teams, and how well validation metrics—particularly map-fitting metrics—performed. Thirteen teams built models for four benchmark maps, three of which formed a resolution series spanning 1.8–3.1 Å. The comparison found that models at near-atomic resolution were generally accurate and reproducible, while different scores captured different aspects of quality. The authors therefore recommend using multiple metrics to provide a comprehensive, objective description of models and how well they fit the observed density. These recommendations apply both to individual experiments and to structural archives such as the PDB. Because the findings are based on the benchmark maps used, achieving the same nominal resolution does not by itself mean that every sample will yield an equally reliable model.
Highlight · The community challenge supports validation with multiple metrics, capturing both reproducibility and support from the experimental map.
DOI: 10.1038/s41592-020-01051-w
05 · Community recommendations on cryoEM data archiving and validation
IUCrJ · Publication / issue date: 2024
Authors: Gerard J. Kleywegt, Paul D. Adams, Sarah J. Butcher, Catherine L. Lawson, Alexis Rohou, Peter B. Rosenthal, Sriram Subramaniam, Maya Topf, Sanja Abbott, Philip R. Baldwin, John M. Berrisford, Gérard Bricogne, Preeti Choudhary, Tristan I. Croll, Radostin Danev, Sai J. Ganesan, Timothy Grant, Aleksandras Gutmanas, Richard Henderson, J. Bernard Heymann, Juha T. Huiskonen, Andrei Istrate, Takayuki Kato, Gabriel C. Lander, Shee-Mei Lok, Steven J. Ludtke, Garib N. Murshudov, Ryan Pye, Grigore D. Pintilie, Jane S. Richardson, Carsten Sachse, Osman Salih, Sjors H. W. Scheres, Gunnar F. Schroeder, Carlos Oscar S. Sorzano, Scott M. Stagg, Zhe Wang, Rangana Warshamanage, John D. Westbrook, Martyn D. Winn, Jasmine Y. Young, Stephen K. Burley, Jeffrey C. Hoch, Genji Kurisu, Kyle Morris, Ardan Patwardhan, Sameer Velankar
Whether cryo-EM structures can be independently reviewed depends on what data are deposited and how maps and models are validated. This paper summarizes a 2020 workshop held at EMBL-EBI, where 47 experts in data processing, model building and refinement, validation, and archiving discussed these issues, with a focus on single-particle analysis. The report describes the workshop background and topics, the consensus recommendations reached, progress in implementing some of them, and challenges for future method development. It is a report on community consensus and implementation progress, not a new algorithm or a single experiment. Since the abstract does not list the archiving requirements item by item, we do not add a checklist here; consult the paper for specific deposition and validation details, and follow the requirements of the relevant database.
Highlight · Community consensus links cryo-EM data archiving with structural validation, making results easier to review independently.
DOI: 10.1107/s2052252524001246
06 · Measurement of atom resolvability in cryo-EM maps with Q-scores
Nature Methods · Publication / issue date: 2020
Authors: Grigore Pintilie, Kaiming Zhang, Zhaoming Su, Shanshan Li, Michael F. Schmid, Wah Chiu
Atoms are not equally resolvable throughout a cryo-EM map, so a single global resolution value cannot capture these differences. The study introduced Q-score, which uses a fitted or map-derived model to quantify how well individual atoms are resolved in the map. It applies to proteins, nucleic acids, water, ligands, and other solvent atoms. Scores can be summarized by residue, nucleotide, or across an entire model; the overall mean is strongly correlated with estimated resolutions for protein and RNA maps. The authors also compared maps of the same protein collected by different laboratories to test how reproducible features from side chains to water and ions are. The metric has one clear prerequisite: the model used must fit the map well. Q-score therefore provides local evidence, but cannot be treated as a fully independent measure of map quality without considering the model's placement.
Highlight · Q-score extends resolvability assessment from the map as a whole to individual atoms, but its interpretation depends on how well the model fits the map.
DOI: 10.1038/s41592-020-0731-1
07 · Fourier shell correlation threshold criteria
Journal of Structural Biology · Publication date: 2005
Authors: Marin van Heel, Michael Schatz
Fourier shell correlation (FSC) is a widely used quality metric for 3D electron microscopy reconstructions, but converting the curve into a resolution value requires choosing a threshold. This paper examines how the number of voxels per shell, structural symmetry, and the fraction of the reconstruction volume occupied by the structure affect FSC statistics. Through theory and model experiments, the authors question the practice of applying a single fixed threshold universally. They compare two types of threshold curves: those based on the sigma factor ask whether the signal is significantly above noise, while those based on bit information ask whether enough interpretable information has been collected. The paper lays bare the statistical assumptions behind resolution values and stresses that thresholds must have a clear meaning. These are the framework and arguments advanced in the paper; they cannot be detached from the specific estimation procedure and used as a replacement for validation standards in all subsequent reconstruction methods.
Highlight · The meaning of an FSC threshold depends on its statistical assumptions: detecting signal and obtaining enough interpretable information are not the same criterion.
DOI: 10.1016/j.jsb.2005.05.009
08 · Optimal Determination of Particle Orientation, Absolute Hand, and Contrast Loss in Single-particle Electron Cryomicroscopy
Journal of Molecular Biology · Publication date: 2003
Authors: Peter B. Rosenthal, Richard Henderson
Single-particle electron microscopy requires not only determining particle orientations but also establishing the absolute hand of the 3D map. The paper uses images of the same particle field taken at two tilt angles, comparing the second image with projections of the map under the known tilt relationship and its inverse to distinguish the correct hand from the incorrect one. The authors also use this independent geometric constraint to optimize orientation searches and refinement, reducing the tendency of a model to match noise in a single image. They validated the method on image pairs of the E2 core of the pyruvate dehydrogenase complex, and recovered high-resolution contrast using noise-correlated structure-factor weighting and a temperature factor. The work links hand determination, orientation reliability, and map contrast; its resolution results come from specific samples and data and should not be taken as a general performance guarantee.
Highlight · Tilt-pair images provide an independent geometric constraint that can check map handedness and help optimize particle orientations.
DOI: 10.1016/j.jmb.2003.07.013
09 · High-resolution noise substitution to measure overfitting and validate resolution in 3D structure determination by single particle electron cryomicroscopy
Ultramicroscopy · Publication date: 2013
Authors: Shaoxia Chen, Greg McMullan, Abdul R. Faruqi, Garib N. Murshudov, Judith M. Short, Sjors H.W. Scheres, Richard Henderson
Iterative single-particle reconstruction can amplify high-frequency noise into apparently reproducible detail, leading to an overestimate of resolution. This study proposes a high-resolution noise-substitution test: Fourier components beyond a specified resolution are replaced with background noise, or their phases are randomized while preserving the corresponding power distribution, and the data are then reconstructed using the same workflow. Comparing FSC curves for the original and noise-substituted data estimates the correlations introduced by overfitting in the final map, allowing resolution estimates to be corrected. The authors tested the method on two sets of β-galactosidase images and with several processing packages, finding that the degree of overfitting depends substantially on the processing workflow. A key advantage is that the method is software-independent and can serve as a parallel control, rather than relying on a more attractive map or a higher FSC alone as evidence of genuine added detail.
Highlight · High-frequency noise-substitution controls can quantify overfitting in reconstruction and provide an independent check on resolution estimates.
DOI: 10.1016/j.ultramic.2013.06.004
10 · Prevention of overfitting in cryo-EM structure determination
Nature Methods · Publication date: 2012
Authors: Sjors H W Scheres, Shaoxia Chen
No abstract was saved in Zotero, and neither the publisher's page nor PubMed provides an accessible abstract. This issue lists metadata only and does not infer the paper's validation workflow from its title.
DOI: 10.1038/nmeth.2115
11 · Quantifying the local resolution of cryo-EM density maps
Nature Methods · Publication date: 2014
Authors: Alp Kucukelbir, Fred J Sigworth, Hemant D Tagare
A single global resolution value cannot capture differences in information across regions of a cryo-EM map. The study defines local resolution in a 3D map using local sinusoidal features and implements the method in ResMap; the abstract states that the algorithm requires no free parameters and can also be applied to other imaging methods, including tomography. The authors analyzed four single-particle reconstructions and sub-tomogram averages, finding that local resolution can range from about 4 to 40 Å. The results highlight that a map can contain both well-resolved and blurred regions, and that the structural information it supports should be interpreted at the relevant scale. The abstract does not specify accuracy limits for all samples, so a particular local estimate should not be taken as direct proof that an atomic model is correct; it is better used to inform local interpretation and further checks.
Highlight · ResMap uses local features to describe variation in map resolution, revealing spatial differences that a global value cannot show.
DOI: 10.1038/nmeth.2727
Atomic models, sequence assignment, and chemical validation (7 papers)
12 · Sequence-assignment validation in cryo-EM models with checkMySequence
Acta Crystallographica Section D Structural Biology · Publication date: 2022
Author: Grzegorz Chojnowski
AI-based structure prediction tools have made cryo-EM modeling easier, but map interpretation, local rebuilding, and refinement can still introduce errors. Register shifts in sequence assignment are especially difficult to spot. This paper introduces checkMySequence, a fast, fully automated method that requires no parameter tuning and identifies mismatches between the residue sequence in a cryo-EM model and its position in the density. The authors show how it can aid modeling when local resolution is too low for visual inspection, and use a much-discussed sequence register error in an RNA polymerase model to illustrate the value of automated checks. The tool targets one specific source of error—sequence register—and complements conventional geometry and map-fit checks. The abstract does not claim that it detects every model problem, so it is best included in a workflow with multiple validation methods rather than used as a single quality criterion.
Highlight · checkMySequence automates the detection of sequence register shifts that are difficult to spot by eye.
DOI: 10.1107/s2059798322005009
13 · Using deep-learning predictions reveals a large number of register errors in PDB depositions
IUCrJ · Publication date: 2024
Authors: Filomeno Sánchez Rodríguez, Adam J. Simpkin, Grzegorz Chojnowski, Ronan M. Keegan, Daniel J. Rigden
Sequence register errors in structural models can escape routine geometry checks and may not noticeably impair map fit. The authors compare residue contacts and distances in experimental models with relationships predicted by methods such as AlphaFold2, flagging regions where the two disagree. This approach is independent of stereochemical and map-model consistency checks and does not directly depend on experimental resolution. A scan of 3–5 Å structures in the PDB identified thousands of possible register errors and proposed candidate corrections; limited implementation of these corrections improved refinement statistics in most cases. The paper also discusses confounding factors, including fold-switching proteins. Disagreement with predictions can therefore prompt a closer look at a model, but a “possible error” still needs to be assessed against the data and does not establish that every flagged region is wrong.
Highlight · Predicted contact relationships provide evidence independent of geometry and map fit, helping locate and correct sequence register errors.
DOI: 10.1107/s2052252524009114
14 · CheckMyMetal: a macromolecular metal-binding validation tool
Acta Crystallographica Section D Structural Biology · Publication date: 2017
Authors: Heping Zheng, David R. Cooper, Przemyslaw J. Porebski, Ivan G. Shabalin, Katarzyna B. Handing, Wladek Minor
Metal ions participate in many biological processes, but their identities and coordination geometries are easy to misassign in structural models. This paper introduces the CheckMyMetal validation service, which detects misidentifications, geometric irregularities, and other implausible features at macromolecular metal-binding sites. The authors organize real examples by metal type to illustrate common problems in crystallographic modeling and offer corresponding modeling and validation recommendations. They note that, at the time, about 40% of macromolecular structures in the PDB contained metal ions, meaning such local errors could affect functional interpretation broadly. CMM turns the chemical plausibility of metal sites into checkable evidence, helping identify positions that need review. Its flags should still be interpreted alongside experimental density, sample conditions, and coordination chemistry; the output alone cannot establish the identity of an unknown ion.
Highlight · CheckMyMetal specifically validates metal identity and coordination geometry, helping uncover site problems that routine checks can miss.
DOI: 10.1107/s2059798317001061
15 · A New Generation of Crystallographic Validation Tools for the Protein Data Bank
Structure · Publication date: 2011
Authors: Randy J. Read, Paul D. Adams, W. Bryan Arendall, Axel T. Brunger, Paul Emsley, Robbie P. Joosten, Gerard J. Kleywegt, Eugene B. Krissinel, Thomas Lütteke, Zbyszek Otwinowski, Anastassis Perrakis, Jane S. Richardson, William H. Sheffler, Janet L. Smith, Ian J. Tickle, Gert Vriend, Peter H. Zwart
The growth of the PDB and mandatory deposition of structure factors created new opportunities to validate models against their underlying diffraction data. This report from the wwPDB X-ray Validation Working Group revisits existing quality standards and stresses that structures can be compared with the database as a whole and with structures at similar resolution, rather than assessed in isolation from a handful of values. The group recommends reporting a small set of easy-to-understand core validation metrics, while retaining details for users who want a deeper analysis, and giving reviewers and editors concise, mature summaries of structural quality. The paper reflects a consensus on validation at the database and peer-review level; it does not propose that any single metric can establish that a structure is entirely correct. Its practical value lies in making quality evidence more transparent and comparable, and helping non-specialists identify structures that warrant closer inspection.
Highlight · Validation reports should provide database-wide and resolution-matched comparisons, alongside a concise set of core quality metrics for reviewers.
DOI: 10.1016/j.str.2011.08.006
16 · MolProbity: More and better reference data for improved all‐atom structure validation
Protein Science · Publication date: 2018
Authors: Christopher J. Williams, Jeffrey J. Headd, Nigel W. Moriarty, Michael G. Prisant, Lizbeth L. Videau, Lindsay N. Deis, Vishal Verma, Daniel A. Keedy, Bradley J. Hintze, Vincent B. Chen, Swati Jain, Steven M. Lewis, W. Bryan Arendall, Jack Snoeyink, Paul D. Adams, Simon C. Lovell, Jane S. Richardson, David C. Richardson
The MolProbity update not only adds checks but also rebuilds the reference data and software infrastructure on which validation depends. This paper describes updated Ramachandran and side-chain rotamer standards, based on roughly a million quality-filtered residues, as well as CaBLAM analysis for checking low-resolution backbones and secondary structure. The program also distinguishes hydrogen distances and van der Waals radii appropriate for X-ray electron-cloud centers versus nuclear positions, and flags rare non-proline cis peptides and twisted peptides. Migration to CCTBX/Python improved maintainability and integration with PHENIX. The authors report that widespread validation and correction have been accompanied by continued improvement in all-atom clashscores for newly submitted structures. These checks primarily assess model plausibility and should still be interpreted alongside the degree of support from experimental data.
Highlight · Improved reference distributions and low-resolution backbone checks help MolProbity pinpoint model outliers in greater detail.
DOI: 10.1002/pro.3330
17 · Mg2+ ions: do they bind to nucleobase nitrogens?
Nucleic Acids Research · Publication date: 2017
Authors: Filip Leonarski, Luigi D'Ascenzo, Pascal Auffinger
The way magnesium ions bind in nucleic acid structures is often used to explain catalytic mechanisms, but the chemical identity of a density peak is not always reliable. The authors examined inner-sphere coordination of Mg2+ to the N1, N3 and N7 atoms of bases in the PDB and found that most magnesium assignments near these nitrogens stem from misinterpreted density; the peaks may instead represent sodium, potassium, ammonium, water or artifacts. With a few documented exceptions, Mg2+ generally does not bind directly to purine N7. Manganese, zinc and cadmium, which have a higher affinity for nitrogen, cannot simply be treated as equivalent to magnesium. The article also discusses potential zinc-binding sites and offers recommendations for ion assignment in crystallography, cryo-EM and modeling. The findings are a reminder that metal-substitution experiments and catalytic interpretations must take coordination chemistry into account; ions should not be assigned based on peak position alone.
Highlight · Density peaks near nucleic acid nitrogens should not automatically be assigned as Mg2+; both ion chemistry and the coordination environment need to be checked.
DOI: 10.1093/nar/gkw1175
18 · Definition of the hydrogen bond (IUPAC Recommendations 2011)
Pure and Applied Chemistry · Publication date: 2011
Authors: Elangannan Arunan, Gautam R. Desiraju, Roger A. Klein, Joanna Sadlej, Steve Scheiner, Ibon Alkorta, David C. Clary, Robert H. Crabtree, Joseph J. Dannenberg, Pavel Hobza, Henrik G. Kjaergaard, Anthony C. Legon, Benedetta Mennucci, David J. Nesbitt
These IUPAC recommendations propose a new definition of the hydrogen bond, drawing on theoretical and experimental evidence accumulated over the past century. The central point is that identifying a hydrogen bond requires supporting evidence, and the recommendations list six types of criteria that can serve as evidence for its existence. The available abstract is very brief and does not describe the criteria individually, so this issue summarizes only the aim of revising the definition and its evidentiary requirements; it does not fill in details from other sources or general knowledge.
Highlight · Hydrogen bonds should be identified on the basis of supporting evidence; consult the original recommendations for the specific criteria.
DOI: 10.1351/pac-rec-10-01-02
Crystallographic Data Quality and Structural Precision (6 papers)
19 · Radiation damage in serial femtosecond crystallography studied in hemoglobin
Acta Crystallographica Section D: Structural Biology · Publication date: 2026-10-01
Authors: M. Galchenkova, I. Dawod, S. Cardoch, J. Sprenger, D. Oberthür, S. Awel, E. De Santis, O. Grånäs, J. Knoska, M. Metz, S. Passmore, G. Pena, I. Sarrou, M. Wiedorn, N. Timneanu, H. Chapman, O. Yefanov, C. Caleman
Whether femtosecond crystallography can capture reliable structures before radiation damage occurs depends on the pulse conditions, not simply on whether the pulses are short enough. At similar peak intensities and a fixed photon energy, the study compared serial diffraction data from hemoglobin crystals collected with 3- and 10-fs pulses. The refined models and data-quality metrics across resolutions were similar for both datasets, and both retained anomalous signal useful for phasing. Combining collision-radiation and molecular-dynamics simulations, the authors found that under the conditions tested, changes in atomic scattering factors were below 1%, and displacements remained below the resolution limit set by photon energy and detector geometry. Together, the experiments and simulations support the conclusion that, within the intensity range studied, 10-fs pulses are sufficient to obtain protein structural information without significant damage. This finding should not be generalized beyond the specific irradiation conditions.
Highlight · In the hemoglobin experiments, 10-fs pulses preserved structural information comparable to that obtained with shorter pulses under the intensities tested.
DOI: 10.1107/S2059798326008569
20 · Polder maps: improving OMIT maps by excluding bulk solvent
Acta Crystallographica Section D: Structural Biology · Publication date: 2017
Authors: Dorothee Liebschner, Pavel V. Afonine, Nigel W. Moriarty, Billy K. Poon, Oleg V. Sobolev, Thomas C. Terwilliger, Paul D. Adams
Crystallographic maps generally incorporate information from the current model, which can introduce model bias during validation. OMIT maps aim to determine whether atoms are supported by the data by removing them from the model and calculating the residual density. But the omitted region may be filled by a flat bulk-solvent model, further obscuring already weak density. Polder maps keep bulk solvent out of the specified OMIT region, improving the display of weak density for ligands, solvent, side chains, alternate conformations, and residues at termini or in loops. The method has been implemented in PHENIX. Its contribution is to address a specific source of bias in map calculation and improve interpretability—not to make the appearance of density automatic proof of molecular identity. Chemical plausibility and other experimental evidence still need to be considered together.
Highlight · By excluding bulk solvent from the OMIT region, Polder maps help prevent the solvent model from masking weak density.
DOI: 10.1107/s2059798316018210
21 · Linking Crystallographic Model and Data Quality
Science · Publication date: 2012
Authors: P. Andrew Karplus, Kay Diederichs
In crystallography, data quality and model fit are often described using metrics on different scales, making it difficult to judge high-resolution cutoffs and how much room remains for model improvement. The article points out that the widely used Rmerge is not a suitable primary basis for setting the high-resolution limit, and that conventional thresholds can discard data that still have value. The authors introduce CC*, which estimates the correlation between observed data and the underlying true signal that cannot be measured directly, allowing model and data quality to be compared on the same scale. This statistical link helps determine which data are worth retaining and whether further model improvement is already limited by data quality. The work provides a statistically grounded basis for decision-making; it does not advocate keeping all weak data indiscriminately. Cutoff choices still need to be assessed in light of the specific data and refinement results.
Highlight · CC* puts model and data quality on a common scale, helping assess the value of weak data and the limits of model improvement.
DOI: 10.1126/science.1218231
22 · How good are my data and what is the resolution?
Acta Crystallographica Section D: Biological Crystallography · Publication date: 2013
Authors: Philip R. Evans, Garib N. Murshudov
After diffraction spots have been integrated, data reduction still involves determining symmetry, scaling observations and assessing data quality. This article describes POINTLESS for symmetry analysis and diagnostic features added to AIMLESS’s scaling model to help determine which measurements are worth retaining. The authors also compare processing statistics, refinement against observed and simulated data, automated model building, and maps at different cutoffs to discuss how to determine the effective resolution. Their tests showed that including weaker high-resolution data beyond commonly used thresholds can bring improvements without causing harm. The key is to judge data by their actual contribution to the model and maps rather than apply a single threshold mechanically. These test results are specific to the data examined and do not mean that data of any quality should always be retained.
Highlight · Resolution cutoffs should be judged by refinement and map quality; weaker data beyond conventional thresholds may still be useful.
DOI: 10.1107/s0907444913000061
23 · Remarks about protein structure precision
Acta Crystallographica Section D: Biological Crystallography · Publication date: 1999
Author: D. W. J. Cruickshank
Structural resolution is related to coordinate precision, but it cannot be taken directly as a measure of uncertainty in every atomic position. Using full-matrix least squares, the article compares restrained and unrestrained refinement, distinguishes the standard uncertainties of atomic positions from those of bond lengths, and uses high-resolution concanavalin A data to demonstrate how strongly precision varies with B factors. Because inverting the full matrix is computationally expensive, the author also reviews several approximation methods and proposes the diffraction-component precision index (DPI), based on R or Rfree, as a rough tool for comparing structures at different resolutions. The paper also highlights how restraints and disordered regions affect interpretation. Its value lies in framing the question of how precise a model really is in terms of uncertainty, rather than assigning one uniform error value to an entire protein.
Highlight · Precision estimates such as DPI help distinguish structural resolution from coordinate uncertainty and the effects of geometric restraints.
DOI: 10.1107/s0907444998012645
24 · Remarks about protein structure precision. Erratum
Acta Crystallographica Section D: Biological Crystallography · Publication date: 1999
Author: D. W. J. Cruickshank
This is an erratum to “Remarks about protein structure precision.” Zotero does not contain an abstract, and no usable abstract was available from the original source. This issue therefore lists only the metadata and does not speculate about the details of the corrections.
DOI: 10.1107/s0907444999004308
Protein Dynamics and Allostery (5 papers)
25 · The ensemble nature of allostery
Nature · Publication date: 2014
Authors: Hesam N. Motlagh, James O. Wrabl, Jing Li, Vincent J. Hilser
Allosteric regulation occurs when a binding event at one site affects a distant functional site. Traditional explanations often focus on a small number of static structures and the transitions between them. Drawing on experimental observations that dynamic and intrinsically disordered proteins can also undergo allostery, this review shifts the focus to conformational ensembles and the statistical properties of interactions. Rather than seek a single structural pathway for transmitting information, it asks how different conformational states and their population distributions jointly encode that information. Analyzing allosteric ensembles reveals diverse regulatory strategies and offers a common framework for comparing different systems. The abstract supports this conceptual perspective and research direction, but does not propose a quantitative model that can be applied directly to every protein; the states, coupling and functional relationships in each system still need to be characterized with experiments and appropriate models.
Highlight · Conformational ensembles and their statistical coupling offer a unified view of allostery in both ordered and disordered proteins.
DOI: 10.1038/nature13001
26 · Dynamic personalities of proteins
Nature · Publication date: 2007
Authors: Katherine Henzler-Wildman, Dorothee Kern
Many proteins have now been structurally characterized, but a static structural snapshot cannot fully explain how they perform their diverse functions. This article emphasizes the dynamic nature of proteins and argues that structural biology needs a time dimension to describe not only where atoms are, but also when they move. The authors frame the goal as observing protein activity in real time at atomic resolution, bringing structural descriptions closer to the processes that underlie function. The abstract primarily sets out this vision and does not specify experimental systems, time resolution or performance results, so we do not present it as evidence that a particular new technology has already achieved the full goal. It offers a conceptual entry point for understanding the relationship between conformational motion and function, and pairs well with this issue’s papers on conformational ensembles, room-temperature structures and ligand-binding pathways.
Highlight · Understanding protein function requires adding a time dimension to static three-dimensional structures to describe motion.
DOI: 10.1038/nature06522
27 · Conformational selection or induced fit: A flux description of reaction mechanism
Proceedings of the National Academy of Sciences · Publication date: 2009
Authors: Gordon G. Hammes, Yu-Chu Chang, Terrence G. Oas
When ligand binding involves a conformational change, simply comparing single-step rate constants cannot establish whether the conformation changes first or the ligand binds first. This article proposes measuring the relative contributions of conformational selection and induced fit by the reaction flux through each pathway, and provides rules for flux analysis of multistep mechanisms. The authors reanalyze two cases: NADPH binding to dihydrofolate reductase, and flavin mononucleotide binding alongside flavodoxin folding. They find that conformational selection predominates at low ligand concentrations, while induced fit predominates at high concentrations; across a broad range of conditions, both pathways carry substantial flux. Thus, a mechanism may not fit neatly into one of two fixed categories, and the balance can shift with experimental conditions. This method offers a quantitative way to interpret kinetic data, but the contributions of particular pathways must still be calculated from data for the system in question.
Highlight · Comparing pathway fluxes, rather than individual rate constants, is key to assessing the relative contributions of conformational selection and induced fit.
DOI: 10.1073/pnas.0907195106
28 · On the nature of allosteric transitions: A plausible model
Journal of Molecular Biology · Publication date: 1965
Authors: Jacques Monod, Jeffries Wyman, Jean-Pierre Changeux
Zotero does not contain an abstract, and attempts to retrieve one from the DOI landing page and PubMed were unsuccessful. This issue lists only the metadata and does not infer the model’s contents from the title.
DOI: 10.1016/s0022-2836(65)80285-6
29 · Accessing protein conformational ensembles using room-temperature X-ray crystallography
Proceedings of the National Academy of Sciences · Publication date: 2011
Authors: James S. Fraser, Henry van den Bedem, Avi J. Samelson, P. Therese Lang, James M. Holton, Nathaniel Echols, Tom Alber
The fact that the overall protein backbone changes little after crystal cooling does not mean that functionally relevant conformational ensembles are fully preserved. Using electron-density sampling, model refinement, and packing-analysis tools, the study compared data from 30 proteins and found that flash cooling can reshape the conformational distributions of more than 35% of side chains and eliminate packing defects associated with functional motions. Analysis of H-Ras also revealed an allosteric network in room-temperature electron density that matched fluctuations observed by solution NMR, but was not apparent in cryogenic maps. These findings suggest that structural databases may be biased toward tighter, more densely packed, apparently single-conformation models. Room-temperature crystallography can therefore add dynamic information relevant to catalysis, ligand binding, and allosteric regulation. The point is that cooling can bias the observed ensemble—not that cryogenic structures have no value.
Highlight · Room-temperature crystallography can reveal conformational ensembles and functional motions that cooling alters or obscures.
DOI: 10.1073/pnas.1111325108
AlphaFold Structure Prediction and Evaluation of Its Applications (5 papers)
30 · Accurate structure prediction of biomolecular interactions with AlphaFold 3
Nature · Publication date: 2024
Authors: Josh Abramson, Jonas Adler, Jack Dunger, Richard Evans, Tim Green, Alexander Pritzel, Olaf Ronneberger, Lindsay Willmore, Andrew J. Ballard, Joshua Bambrick, Sebastian W. Bodenstein, David A. Evans, Chia-Chun Hung, Michael O’Neill, David Reiman, Kathryn Tunyasuvunakool, Zachary Wu, Akvilė Žemgulytė, Eirini Arvaniti, Charles Beattie, Ottavia Bertolli, Alex Bridgland, Alexey Cherepanov, Miles Congreve, Alexander I. Cowen-Rivers, Andrew Cowie, Michael Figurnov, Fabian B. Fuchs, Hannah Gladman, Rishub Jain, Yousuf A. Khan, Caroline M. R. Low, Kuba Perlin, Anna Potapenko, Pascal Savy, Sukhdeep Singh, Adrian Stecula, Ashok Thillaisundaram, Catherine Tong, Sergei Yakneen, Ellen D. Zhong, Michal Zielinski, Augustin Žídek, Victor Bapst, Pushmeet Kohli, Max Jaderberg, Demis Hassabis, John M. Jumper
AlphaFold 3 expands the scope of structure prediction beyond protein-centered systems to a much broader range of biomolecular complexes. Its substantially updated diffusion architecture predicts structures jointly for proteins, nucleic acids, small molecules, ions, and modified residues. On the benchmarks used in the paper, the authors report better protein–ligand predictions than leading docking tools, better protein–nucleic acid interaction predictions than specialized predictors, and better antibody–antigen predictions than AlphaFold-Multimer v2.3. A major advance is bringing several interaction types that were previously handled separately into a unified deep-learning framework. These comparisons apply to the specific datasets and tasks tested; they do not mean that every complex system can be predicted reliably. Particular binding modes still need to be assessed alongside confidence scores and independent experimental evidence.
Highlight · Its diffusion architecture lets AlphaFold 3 predict structures involving diverse biomolecules in a unified framework.
DOI: 10.1038/s41586-024-07487-w
31 · Addendum: Accurate structure prediction of biomolecular interactions with AlphaFold 3
Nature · Publication date: 2024
Authors: Josh Abramson, Jonas Adler, Jack Dunger, Richard Evans, Tim Green, Alexander Pritzel, Olaf Ronneberger, Lindsay Willmore, Andrew J. Ballard, Joshua Bambrick, Sebastian W. Bodenstein, David A. Evans, Chia-Chun Hung, Michael O’Neill, David Reiman, Kathryn Tunyasuvakool, Zachary Wu, Akvilė Žemgulytė, Eirini Arvaniti, Charles Beattie, Ottavia Bertolli, Alex Bridgland, Alexey Cherepanov, Miles Congreve, Alexander I. Cowen-Rivers, Andrew Cowie, Michael Figurnov, Fabian B. Fuchs, Hannah Gladman, Rishub Jain, Yousuf A. Khan, Caroline M. R. Low, Kuba Perlin, Anna Potapenko, Pascal Savy, Sukhdeep Singh, Adrian Stecula, Ashok Thillaisundaram, Catherine Tong, Sergei Yakneen, Ellen D. Zhong, Michal Zielinski, Augustin Žídek, Victor Bapst, Pushmeet Kohli, Max Jaderberg, Demis Hassabis, John M. Jumper
This is an addendum to the original AlphaFold 3 paper. Zotero has no abstract saved, and the original publication page provides no usable abstract. This issue therefore lists the metadata only and does not speculate about the addendum’s contents.
DOI: 10.1038/s41586-024-08416-7
32 · AlphaFold predictions are valuable hypotheses and accelerate but do not replace experimental structure determination
Nature Methods · Publication date: 2024
Authors: Thomas C. Terwilliger, Dorothee Liebschner, Tristan I. Croll, Christopher J. Williams, Airlie J. McCoy, Billy K. Poon, Pavel V. Afonine, Robert D. Oeffner, Jane S. Richardson, Randy J. Read, Paul D. Adams
Whether high confidence in an AlphaFold model guarantees that its local details are correct is a key question when using predicted structures. The authors directly compared predicted models with experimental crystallographic maps and found that many predictions agree closely with experiment, but some high-confidence models still show overall distortions, incorrect domain orientations, or mismatches in local backbone and side-chain conformations. They also note that the predictions examined did not account for ligands, covalent modifications, or other environmental factors. The authors recommend treating predictions as highly valuable structural hypotheses and interpreting them in light of confidence scores, with experimental validation especially important for interactions and key details not included in the prediction. The study does not discount the role of predictions in accelerating structure determination; rather, it clarifies the boundary between a useful hypothesis and experimentally verified detail.
Highlight · Even high-confidence predictions can miss experimental details; AlphaFold models are best treated as structural hypotheses to be validated.
DOI: 10.1038/s41592-023-02087-4
33 · Highly accurate protein structure prediction with AlphaFold
Nature · Publication date: 2021
Authors: John Jumper, Richard Evans, Alexander Pritzel, Tim Green, Michael Figurnov, Olaf Ronneberger, Kathryn Tunyasuvakool, Russ Bates, Augustin Žídek, Anna Potapenko, Alex Bridgland, Clemens Meyer, Simon A. A. Kohl, Andrew J. Ballard, Andrew Cowie, Bernardino Romera-Paredes, Stanislav Nikolov, Rishub Jain, Jonas Adler, Trevor Back, Stig Petersen, David Reiman, Ellen Clancy, Michal Zielinski, Martin Steinegger, Michalina Pacholska, Tamas Berghammer, Sebastian Bodenstein, David Silver, Oriol Vinyals, Andrew W. Senior, Koray Kavukcuoglu, Pushmeet Kohli, Demis Hassabis
The number of known protein sequences far exceeds the number of experimentally determined structures, making reliable sequence-to-structure prediction fundamental. This paper introduces a redesigned AlphaFold neural network. In the CASP14 blind assessment, it achieved accuracy competitive with experimental structures for most targets, including high accuracy when no similar known structure was available. The method incorporates physical and biological knowledge of protein structure, along with information from multiple sequence alignments, into a deep-learning architecture, reducing reliance on existing homologous templates. This work greatly broadened access to structural research and laid the groundwork for large-scale structural bioinformatics. The abstract reports overall performance in a blind assessment; it does not mean that every residue, state, or arbitrary complex reaches experimental accuracy. The uncertainty of each prediction still needs to be evaluated.
Highlight · AlphaFold demonstrated at CASP14 that it can predict highly accurate structures from sequence, even without relying on homologous templates.
DOI: 10.1038/s41586-021-03819-2
34 · A structural biology community assessment of AlphaFold2 applications
Nature Structural & Molecular Biology · Publication date: 2022
Authors: Mehmet Akdel, Douglas E. V. Pires, Eduard Porta Pardo, Jürgen Jänes, Arthur O. Zalevsky, Bálint Mészáros, Patrick Bryant, Lydia L. Good, Roman A. Laskowski, Gabriele Pozzati, Aditi Shenoy, Wensi Zhu, Petras Kundrotas, Victoria Ruiz Serra, Carlos H. M. Rodrigues, Alistair S. Dunham, David Burke, Neera Borkakoti, Sameer Velankar, Adam Frost, Jérôme Basquin, Kresten Lindorff-Larsen, Alex Bateman, Andrey V. Kajava, Alfonso Valencia, Sergey Ovchinnikov, Janani Durairaj, David B. Ascher, Janet M. Thornton, Norman E. Davey, Amelie Stein, Arne Elofsson, Tristan I. Croll, Pedro Beltrao
A prediction that closely resembles an experimental structure is not automatically suitable for every structural biology task. This community assessment examines AlphaFold2’s applications in identifying structural features, interpreting missense variants, predicting function and ligand-binding sites, modeling interactions, and building models from experimental data. Analysis of 11 proteomes found that, on average, about 25% more residues could be modeled with confidence than with comparative modeling, and revealed structural features rarely seen in the PDB. The study also assesses disordered regions and complex predictions, concluding that AF2 models can be as useful as experimental models for many tasks, provided confidence metrics are interpreted critically. The focus is on evaluating real-world applications and their limitations—not on treating all predictions as substitutes for experimental structures.
Highlight · AF2 expands the range of structures that can be modeled with confidence, but its confidence metrics still need careful interpretation across tasks.
DOI: 10.1038/s41594-022-00849-w
Membrane Receptors and Ion Channels (4 papers)
35 · Molecular Pharmacology of Dopamine Receptors: Structure, Function, and Therapeutic Implications
Annual Review of Pharmacology and Toxicology · Publication date: 2026
Authors: Youwen Zhuang, H. Eric Xu
Dopamine receptors have long been targets for drug development, but available treatments remain limited by inadequate subtype selectivity, insufficient specificity in pathway engagement, and adverse effects. Using active and inactive receptor structures as a framework, this review brings together structural, biochemical, pharmacological, and computational studies to discuss activation, inverse agonism, transducer coupling, biased signaling, allosteric modulation, and ligand selectivity. The authors focus in particular on how ligand-induced conformations, transducer interactions, and the membrane environment jointly shape signaling outputs, and compare activation pathways in D1-like and D2-like receptors. The review also considers what allosteric sites and AI-assisted discovery may mean for drug design. Its aim is to move from descriptive pharmacology toward a mechanistic, structure-guided design framework—not to claim that a new therapy has already improved clinical safety or efficacy.
Highlight · Drug selectivity at dopamine receptors must be understood in terms of conformation, coupling partners, and the membrane environment.
DOI: 10.1146/annurev-pharmtox-060425-070659
36 · Structural basis of voltage gating in a partially activated K+ channel
Nature Communications · Published/issue date: 2026-10-01
Authors: Richa Agrawal, Ramon Mendoza Uriarte, Bernardo I. Pinto-Anwandter, Trayder Thomas, Lydia Blachowicz, Young Hoon Koh, Francisco Bezanilla, Eduardo Perozo, Benoît Roux
How voltage-gated potassium channels convert movement of the voltage sensor into pore opening is a central question in electromechanical coupling. The authors determined the cryo-EM structure of the Shaker channel ILT mutant, which partially uncouples gating-charge movement in the voltage-sensing domain from pore opening. This captured an intermediate state in which the sensor is partly activated while the pore gate remains closed. Combined with AlphaFold2-based modeling and molecular dynamics simulations, the study proposes that shifts in the population balance between two metastable positions of the S4–S5 linker transmit sensor activation to the pore. This result brings dynamic conformational distributions into the coupling mechanism, rather than relying solely on comparisons between two endpoint structures. The evidence comes from a specific mutant system and computational analyses; its generality to other channels requires further validation.
Highlight · Shifts between metastable states of the S4–S5 linker offer a dynamic explanation for coupling between the voltage sensor and pore.
DOI: 10.1038/s41467-026-78089-5
37 · GPCR antagonism via rewiring of receptor trafficking and degradation
Nature · Published/issue date: 2026-09-30
Authors: Kaitlin Rhee, Lawrence Shue, Akimasa Adachi, Pengwei Sun, James Osei-Owusu, Dingjingyu Zhou, Aoxing Cheng, Yi Ran Xu, Qingyue Li, Apoorva Baluapuri, Edward P. Harvey, Karen Adelman, Meredith A. Skiba, Bianxiao Cui, Jun R. Huh, Andrew C. Kruse, Xin Zhou
Conventional GPCR antagonists mainly block receptor activity; this study instead redirects receptor trafficking and degradation. The authors engineered a bispecific antibody chimera, GTAC, that recognizes both GPCRs and the transferrin receptor TfR1, prompting target-receptor internalization and routing it to lysosomal degradation. Across tested targets including BILF1, RXFP1, and CCR6, GTAC selectively reduced receptor levels and suppressed receptor output, including constitutive signaling. Its reported potency was more than one to two orders of magnitude greater than that of conventional antibody antagonists. Protein engineering and multicolor live-cell imaging further clarified the cellular mechanism and context-dependent design requirements. The work shows that altering receptor fate can provide a functional antagonism strategy, but the reported potency comes from experimental systems and cannot be equated directly with therapeutic benefit in humans.
Highlight · Routing GPCRs to internalization and lysosomal degradation enables functional antagonism by reducing receptor abundance.
DOI: 10.1038/s41586-026-11088-0
38 · Structural basis of stepwise G protein activation by the viral chemokine receptor US28
Nature Communications · Published/issue date: 2026-09-28
Authors: Kevin M. Jude, Carl-Mikael Suomivuori, Deepa Waghray, Shoji Maeda, Yoshinori Fujiyoshi, Asuka Inoue, K. Christopher Garcia, Naotaka Tsutsumi
The process by which GPCRs recognize G proteins and trigger GDP release is difficult to reconstruct from a single stable complex structure. Using the viral chemokine receptor US28, the researchers obtained cryo-EM structures of three US28–Gq states: a GDP-bound encounter complex, the canonical nucleotide-free state, and a possible intermediate linking the two. Together with simulations and functional data, these structures support a conformational model for stepwise G-protein activation and echo previously proposed mechanisms for human GPCR activation. The work provides several high-resolution snapshots that support a possible trajectory; it does not directly observe the entire activation process continuously. Its main significance is to connect structural changes from initial recognition to subsequent nucleotide release, providing a basis for comparing receptor signaling mechanisms.
Highlight · Multiple US28–Gq structures outline a stepwise model of G-protein activation, from initial contact to nucleotide release.
DOI: 10.1038/s41467-026-78122-7
Brain activity and mechanisms of neurological disorders (3 papers)
39 · Psychedelics align brain activity with context
Nature · Published/issue date: 2026-08
Authors: Devon Stoliker, Leonardo Novelli, Moein Khajehnejad, Mana Biabani, Matthew D. Greaves, Tamrin Barta, Martin Williams, Sidhant Chopra, Olivier Bazin, Otto Simonsson, Richard Chambers, Frederick S. Barrett, Gustavo Deco, Katrin H. Preller, Robin L. Carhart-Harris, Anil K. Seth, Suresh Sundram, Gary F. Egan, Adeel Razi
Brain activity under psychedelics is often described as more disordered, but time-averaged measures may miss its underlying organization. The study recorded functional MRI and EEG from 62 adults, comparing activity before dosing with activity on the day of psilocybin administration during rest, meditation, music, and film. Using machine learning to represent each person's brain dynamics as low-dimensional trajectories, the authors found that activity under the drug still showed structured patterns aligned with context; networks typically associated with internal and external processing were more integrated. This degree of contextual alignment was associated with ego dissolution, feelings of merging with the environment, and changes in mood the following day. The findings offer a dynamic account of apparent “disorder,” but correlations between experience and brain activity do not establish long-term therapeutic effects or a single causal mechanism.
Highlight · Low-dimensional trajectories reveal a hidden order in brain activity under psilocybin, structured by context.
DOI: 10.1038/s41586-026-10910-z
40 · Circuit mechanisms of psychotic symptoms across disorders
Neuron · Published/issue date: 2026/10/01
Authors: Urs Braun, Johannes Wolf, Dusan Hirjak, Emanuel Schwarz, Heike Tost, Andreas Meyer-Lindenberg
Hallucinations, delusions, and disorganized thinking occur across a range of psychiatric and neurological disorders, yet there is no unified explanation of how molecular abnormalities give rise to symptoms. This article proposes a transdiagnostic framework: disruptions in dopamine, glutamate, GABA, serotonin, and acetylcholine alter neuronal gain and synaptic efficacy, affecting the precision weighting of sensory evidence, prediction errors, and internal representations. These local changes propagate through cortico-striato-thalamo-cortical circuits, leading to abnormal integration and segregation in large-scale networks and unstable attractor states. The model allows different biological pathways to converge on shared computational vulnerabilities and proposes individualized testing using imaging, dynamical-systems methods, and AI. This is a mechanistic synthesis and a testable framework, not a validated diagnostic or treatment tool.
Highlight · A cross-scale framework links neurotransmitter disruptions and circuit dynamics to psychotic symptoms shared across disorders.
DOI: 10.1016/j.neuron.2026.09.011
41 · Prion-like transmission of human tau strains in the mouse brain
Nature · Published/issue date: 2026-09-30
Authors: Sofia Lövestam, Aki Shimozawa, Airi Tarutani, Reiko Ohtani, Masami Masuda-Suzukake, Kazuko Hasegawa, Andrew C. Robinson, Yuko Saito, Shigeo Murayama, Mari Yoshida, Hisaomi Suzuki, Mitsumoto Onaya, Masato Hasegawa, Michel Goedert, Sjors H. W. Scheres
Neurodegeneration-associated fibrils are thought to spread through templated seeding, but direct evidence is needed to determine whether distinct “strains” retain their original structures after transmission. The researchers injected tau fibrils from the brains of people with Alzheimer’s disease or corticobasal degeneration into the brains of wild-type mice. Newly formed amyloid fibrils made of mouse tau were found to share the same structures as the human-derived seeds. The results support the idea that tau folds can be replicated by templating and provide structural evidence that distinct conformations correspond to different pathological types. The authors propose that mice can be used to study the molecular mechanisms by which specific tau folds drive pathology. Here, “prion-like” describes templated conformational transmission in an experiment; it does not imply that these neurodegenerative diseases are contagious through human contact.
Highlight · Human tau seeds induced the same fold in mice, supporting templated transmission of strain-specific structural identity.
DOI: 10.1038/s41586-026-11061-x
Protein structure alignment (2 papers)
42 · A normalized root‐mean‐spuare distance for comparing protein three‐dimensional structures
Protein Science · Published/issue date: 2001
Authors: Oliviero Carugo, Sándor Pongor
Root-mean-square deviation (RMSD) is a widely used measure for comparing protein structures, but its value also depends on the number of equivalent atom pairs included in the alignment. As a result, two structure pairs of markedly different lengths may have the same RMSD without being equally similar. This paper proposes a simple normalization procedure intended to reduce this size dependence and make comparisons across structures of different sizes more comparable. The authors suggest that the approach could be used in studies of structural evolution, fold classification, and direct comparison of structural models. The abstract does not provide the formula or range of applicability, so we do not add operational details here. The paper serves as a reminder that when citing a single measure of structural difference, the alignment size and normalization method should also be specified rather than ranking all RMSD values on the same scale.
Highlight · Structural comparisons need to account for alignment size; normalized RMSD aims to reduce size-related bias.
DOI: 10.1110/ps.690101
43 · TM-align: a protein structure alignment algorithm based on the TM-score
Nucleic Acids Research · Published/issue date: 2005
Authors: Yang Zhang, Jeffrey Skolnick
TM-align combines a TM-score rotation matrix with dynamic programming to find structural alignments between two protein chains. In the paper’s benchmarks, it was faster than CE, DALI, and SAL, and performed better in average alignment accuracy and coverage. The authors also conducted pairwise comparisons of more than 10,000 representative PDB chains, classified structural folds at a specified TM-score threshold, and tested the similarity of predicted models to known structures. One key observation was that both correctly and incorrectly folded models could have close structural analogs, so finding a similar structure alone is not proof that a prediction is correct. However, similarity still correlates with model accuracy in ways that can help with model selection. The tool’s value lies in efficiently quantifying structural similarity, while interpretation depends on the alignment range and intended use.
Highlight · TM-align efficiently measures structural similarity, while reminding us that finding a similar structure does not prove a model is correct.
DOI: 10.1093/nar/gki524
Crystals and biological assemblies (2 papers)
44 · Inference of Macromolecular Assemblies from Crystalline State
Journal of Molecular Biology · Published/issue date: 2007
Authors: Evgeny Krissinel, Kim Henrick
Symmetry contacts in crystal structures may reflect genuine biological assemblies or merely crystal packing, and distinguishing between the two matters for functional interpretation. This paper discusses the physicochemical principles underlying stable macromolecular complexes and methods for calculating affinity and assembly entropy, the latter of which has a major effect on complex size and symmetry. On this basis, the authors developed a method grounded in chemical thermodynamics to automatically identify likely assemblies from PDB entries derived from X-ray diffraction, and implemented it as a public web service. The paper reports a success rate of about 80–90% for recovering biological units in its evaluation. This provides a tool for studying protein interactions using crystal structures, while also underscoring the uncertainty in such inferences: particular assemblies, especially those involving weak interactions, should still be assessed alongside other evidence.
Highlight · Thermodynamics-based assembly inference helps distinguish biological complexes from crystal packing, but predictions are not certain in every case.
DOI: 10.1016/j.jmb.2007.05.022
45 · Macromolecular complexes in crystals and solutions
Acta Crystallographica Section D Biological Crystallography · Published/issue date: 2011
Author: Evgeny Krissinel
Contacts in a crystal do not always faithfully reflect biological assemblies in solution. This paper reviews established approaches to analyzing macromolecular interactions and complexes from crystal packing, and identifies conditions under which routine inference can fail. Computational analyses indicate that weak interactions with dissociation constants of around 100 μM or higher are likely to be lost during crystallization. This means the risk of disagreement between crystal structures and native assemblies cannot be ignored for some PDB dimers. The author estimates that for about one-fifth of protein dimers, the probability of such a discrepancy exceeds 50%, and recommends supplementing crystal-based inference with non-crystallographic studies when assessing weakly bound complexes. These percentages are computational estimates from the paper; they are best understood as a caution about methodological limits, not as a direct verdict that any particular database assembly is wrong.
Highlight · Weakly bound complexes may change during crystallization, so crystal-based assembly predictions should be supported by solution studies or other independent evidence.
DOI: 10.1107/s0907444911007232
Other papers (6 papers)
46 · Bluues2: An enhanced web server for fast and accurate electrostatic analysis of proteins and protein complexes
Protein Science · Published/issue date: 2026
Authors: Damiano Piovesan, Nima Rasekh, Miguel Angel Soler, Walter Rocchia, Gennaro Esposito, Federico Fogolari, Silvio C. E. Tosatto
Electrostatic interactions affect protein stability, binding, and function, but analyzing complexes often means switching between computational and visualization tools. Bluues2 updates a service for protein electrostatics based on the generalized Born model. It uses NanoShaper to generate molecular surfaces, improves GB radius calculations, and extends analysis to protein–protein and protein–ligand complexes. The service calculates pKa, surface potential, solvation free energy, binding free energy, and interfacial electrostatic complementarity, while integrating the Mol* viewer and DRMAAtic task management. Its contribution is to bring these analyses together in a single interface, making it easier to interpret structures and interface properties. The abstract provides no quantitative benchmarks for improved accuracy, however, so it does not establish that the service outperforms other methods across all systems.
Highlight · Bluues2 brings a range of electrostatic analyses for proteins and complexes together in a single visualization service.
DOI: 10.1002/pro.70815
47 · The gasdermin family: from pyroptosis mechanisms to therapeutic targets
Signal Transduction and Targeted Therapy · Published: October 2, 2026
Authors: Weilv Xu, Shiyang Liu, Zexu Yu, Jinhuang Shi, Yang Yang, Fushan Shi
Gasdermins are key executors of pyroptosis, but their roles extend beyond inflammatory membrane rupture. This review surveys how family members are activated and regulated, how they form pores and drive cell rupture, and how membranes are repaired. It also covers functions that may or may not depend on pyroptosis, including mucus secretion, bone resorption, and mitochondrial metabolism. The authors discuss crosstalk with apoptosis, necroptosis, NETosis, and PANoptosis, as well as context-dependent roles in host defense, chronic inflammation, and tumor immunity. The therapeutic section reviews small molecules such as disulfiram and material-based delivery strategies. The review emphasizes that this family can both protect the host and drive disease, so interventions must account for tissue and disease context. The strategies discussed reflect research progress and potential, not therapies that have all been clinically validated.
Highlight · Understanding the therapeutic potential of gasdermins requires accounting for functions beyond pyroptosis and their context-dependent effects.
DOI: 10.1038/s41392-026-02867-2
48 · Function-preserving watermarking of AI-generated proteins
Nature · Published: September 30, 2026
Authors: David Stutz, Alexander I. Cowen-Rivers, Guillermo Ortiz-Jimenez, Jeremy Ratcliff, Vinicius Zambaldi, Lindsay Willmore, Josh Abramson, Harshnira Patani, Christina Kouridi, Florian Stimberg, Mel Vecerik, Alex Chu, Sukhdeep Singh, Sumanth Dathathri, Eliseo Papa, Valentin De Bortoli, Arnaud Doucet, Demis Hassabis, Jue Wang, Sven Gowal, Pushmeet Kohli
Generative AI is accelerating protein design, but it also makes it harder to trace the origins of sequences and structures. The authors introduce SynthIDBio, which embeds detectable watermarks in both protein sequences and predicted structures. The sequence approach adds an identifier while preserving function: experimentally designed binding proteins showed binding affinities similar to those of unwatermarked controls, and watermark detection accuracy was nearly perfect in tests. The structural approach fine-tunes AlphaFold3 to embed subtle watermarks in biomolecular structures. By treating traceability and function preservation as joint design goals, the study provides a proof of concept for tracking the provenance of AI-engineered biological products. The findings do not establish that every protein or downstream transformation has been tested, nor that watermarks cannot be removed by arbitrary modifications.
Highlight · SynthIDBio demonstrates the feasibility of watermarking AI-generated protein sequences and structures without compromising function.
DOI: 10.1038/s41586-026-10965-y
49 · Soft drugs: design principles and topical applications
Nature Reviews Drug Discovery · Published: September 30, 2026
Authors: Tracey Pirali, Alessandro Accetta, Jarkko Rautio, Laura Carzaniga
Soft-drug design aims for a drug to exert its effects first, then become inactive rapidly through a predictable metabolic pathway, helping control duration of action and systemic exposure. This review revisits the design principles and distinguishes soft drugs from prodrugs, which require metabolic activation. For systemic use, rapidly waning effects can be useful in settings such as anesthesia; for topical use, the goal is to retain efficacy at the site of administration while limiting toxicity after systemic absorption. The authors review clinical progress in dermatological, inhaled, and gut-restricted therapies, and discuss practical ways to incorporate soft-drug principles into drug discovery. They argue that the strategy remains underused and is often confused with other concepts. Its value lies in making metabolic inactivation an intentional design goal—not in assuming that faster metabolism is always better. Safety and efficacy still depend on the drug and how it is administered.
Highlight · Soft drugs are designed to become predictably inactive after acting, balancing local efficacy, duration of action, and systemic exposure.
DOI: 10.1038/s41573-026-01535-y
50 · Uniform thin ice on ultraflat graphene for high-resolution cryo-EM
Nature Methods · Published: January 2023
Authors: Liming Zheng, Nan Liu, Xiaoyin Gao, Wenqing Zhu, Kun Liu, Cang Wu, Rui Yan, Jincan Zhang, Xin Gao, Yating Yao, Bing Deng, Jie Xu, Ye Lu, Zhongmin Liu, Mengsen Li, Xiaoding Wei, Hong-Wei Wang, Hailin Peng
Uneven ice thickness during cryo-EM sample preparation can compromise image quality, particularly limiting high-resolution analysis of small-molecular-weight samples. The study found that uniform thin ice is linked to the flatness of the support film and used ultraflat graphene as a support to improve control of vitreous ice thickness. Using this method, the authors determined structures of 64-kDa hemoglobin, asymmetric 67-kDa alpha-fetoprotein, and 52-kDa streptavidin at resolutions of 3.5, 2.6, and 2.2 Å, respectively. The results show that support flatness can be a key variable in optimizing sample preparation. The paper also points to potential applications in cryo-electron tomography and structure-guided drug discovery, but these prospects should be distinguished from the demonstrated single-particle results for small proteins; not all samples can be assumed to benefit equally.
Highlight · Ultraflat graphene improves thin-ice uniformity, helping produce high-quality cryo-EM structures of several low-molecular-weight samples.
DOI: 10.1038/s41592-022-01693-y
51 · In-cell discovery and characterization of a non-canonical bacterial protein translocation-folding complex
Cell · Published: September 25, 2026
Authors: Rasmus K. Jensen, Liang Xue, Federico Marotta, Joseph C. Somody, Joel Selkrig, Swantje Lenz, Juri Rappsilber, Mikhail M. Savitski, Jan Kosinski, Athanasios Typas, Maria Zimmermann-Kogadeeva, Peer Bork, Julia Mahamid
In-cell structural imaging can reveal previously unknown complexes, but establishing their composition and function requires multiple lines of evidence. Using genome-reduced Mycoplasma pneumoniae as a model, the researchers mapped an unknown complex on the cell surface in situ by cryo-electron tomography. They combined this with proteomics, structure prediction, integrative modeling, and bioinformatics to identify a conserved Sec translocation system paired with an extracellular dome-like structure. The latter consists mainly of three previously uncharacterized Mdp proteins and is structurally homologous to the ATP-independent foldase PrsA; Mdp444 retains key catalytic residues and activity. The subnanometer-resolution map links SecA-mediated cotranslational translocation to an extracellular protein-folding system, demonstrating how in situ structural analysis combined with molecular identification can help explain unknown cellular machinery.
Highlight · In situ tomography, combined with omics and prediction, links the bacterial Sec translocation machinery to an extracellular protein-folding system.
DOI: 10.1016/j.cell.2026.08.055