Both studies emerged from the group of Julia Mahamid, an EMBL group leader and senior author on the two papers. Her team harnessed cryo-electron tomography, or cryo-ET, an imaging technique that flash-freezes cells and then images them in three dimensions at near-atomic resolution. Unlike traditional structural biology methods, which require molecules to be extracted and studied outside their cellular context, cryo-ET allows scientists to peer directly into intact, frozen cells and visualize the actual arrangements of molecular machines as they exist in life. The result is a view of the cell that preserves the spatial relationships, physical contacts, and functional cross-talk that conventional approaches inevitably destroy.

The first of the two papers, led by Joe Dobbs, a former PhD fellow in the Mahamid group who is now a postdoctoral researcher at the Max Planck Institute for Brain Research in Frankfurt, focused on Mycoplasma pneumoniae, a bacterium with one of the smallest and simplest genomes known. This minimal cell is an ideal testing ground for a new kind of analysis: with fewer components and less redundancy, every molecular interaction stands out more clearly. Dobbs and his colleagues used cryo-ET to reconstruct high-resolution maps of the bacterium’s interior and to count instances of ribosomes—the protein-producing factories of the cell—in different functional states across hundreds of individual cells. Because proteins carry out nearly every task a cell needs to survive, understanding how they are made is fundamental to understanding how cells work, why diseases arise, and how better medicines can be designed.

What the team found went well beyond a simple census of cellular factories. Among their observations were several previously unseen complexes that physically connect transcription, the process of producing messenger RNA from DNA, with translation, the process by which ribosomes read that RNA and build proteins. These direct links provided structural evidence for so-called supercomplexes—multi-machine assemblies that researchers had long hypothesized but never directly visualized in this context. The findings offer new molecular mechanisms for how these intertwined processes are coordinated and regulated, suggesting that the cell does not run its production line as a series of disconnected steps but as a tightly coupled system in which one stage feeds directly into the next.

The study also produced a surprise at the cell’s periphery. The researchers found parts of ribosomes attached to the inner face of the cell membrane even when those subunits were not actively producing proteins. Their results suggest that these membrane-bound subunits may detach only when conditions allow a new round of protein synthesis to begin. Remarkably, similar behavior was observed in mammalian cells decades ago, and the new bacterial findings imply that this regulatory mechanism may be conserved across the enormous evolutionary distance separating bacteria from humans. In other words, a principle of cellular organization first glimpsed in our own cells may turn out to be an ancient, universal feature of life.

“I think of it as a city,” Dobbs explained. “The molecular machines I work on are like little factories that produce things, while some of my colleagues have been studying systems that transport and process those products. Together, they form different parts of a living molecular community inside the cell – much like a city with different specialised activities all working together.” It is precisely this focus on interactions between molecular machines, rather than on the machines in isolation, that distinguishes the two papers from the bulk of prior structural biology and marks them as proofs of concept for a new analytical paradigm.

The second paper, led by Rasmus Jensen of the Mahamid group, approached the cell from the opposite direction. Rather than starting with a specific biological question, Jensen began with an observation: something unusual and unexplained was visible in the tomographic images of the cell. “Joe started with a biological question he wanted to answer,” Jensen said. “I started from the opposite direction. We’d seen something interesting in the cell and worked backwards to figure out what it was, eventually discovering that it’s a new configuration of a molecular machine that helps proteins either be exported out of the cell or inserted into its membrane.” That reverse-engineering approach led the team to a discovery that no one had anticipated: an entirely new protein-folding machinery operating at the cell surface.

By combining cryo-ET imaging with proteomics—the large-scale identification of the proteins present in a sample—and computational protein structure prediction, the researchers determined that this newly characterized machinery contains the well-known Sec-translocon, a protein channel embedded in the cell membrane that functions as a doorway through which newly made proteins exit the cell or become inserted into the membrane itself. Alongside the translocon, the team identified three previously unknown proteins that assist transported proteins in folding into their correct functional shapes after they emerge at the cell surface. The cryo-ET data yielded the most detailed images to date of the conserved portion of the bacterial protein transport machinery, showing how proteins fresh from the ribosome begin their journey through the membrane and revealing that transport is closely linked to a folding system operating outside the cell—a location where such machinery had not been described before in this organism.

“Working together with many colleagues at EMBL, we combined expertise in microbiology, proteomics, bioinformatics, structural biology, and integrative modelling,” Mahamid said. “This paper tells the entire story of the discovery process: visualising a new cellular machinery, identifying the genes and proteins, and figuring out how it all functions. That’s a remarkable achievement and opens new ways to identify new complexes and functions in different systems.” The workflow demonstrated in the study—imaging an unknown structure, tracing it back to its genetic basis, and then characterizing its function—offers a template that other laboratories can apply to organisms far more complex than a minimal bacterium.

Indeed, the broader significance of both studies lies less in any single finding than in the conceptual foundation they establish. “In both Joe’s work and mine, we’ve started with what we’d call a minimal cell where everything is relatively simple,” Jensen noted. “We’ve advanced what is conceptually possible in this minimal bacterium and established ways to apply these methods to more complex organisms.” Mahamid frames the future in similar terms: “The future is studying molecular communities. Decades of structural biology done on isolated molecules provided us with detailed views of their structures and functions. In the past, we just didn’t have the technology to study their more complex interactions and cross-talk. This work points again to how cryo-ET is an enabling technology, allowing us to see new things and connectivity inside cells. Its ability to show us more seems only limited by our asking the right questions.” As cryo-ET continues to mature, the molecular cities inside cells—bacterial and human alike—are finally beginning to reveal not just their buildings, but the bustling traffic that connects them.

Subject of Research: Visualizing interconnected protein production, transport, and folding machineries in intact bacterial cells using cryo-electron tomography

Article Title: The bustling molecular communities inside cells

Article References: The bustling molecular communities inside cells. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: cryo-electron tomography, Mycoplasma pneumoniae, ribosomes, transcription, translation, Sec-translocon, protein folding, protein transport, structural biology, supercomplexes, cell membrane, EMBL

One of the most striking threads in the collection concerns the way human expertise quietly moderates algorithmic decisions in high-stakes environments. Carboni and colleagues examine clinical settings where violence is a real possibility, showing how nurses intervene in and influence algorithmic outputs. The researchers highlight the importance of the doubt that nurses introduce into the decision-making process, treating this cultivated skepticism not as friction to be eliminated but as a vital contribution to collective judgment. In a related vein, Van Voorst documents what is called collaborative tinkering between humans and algorithms in the context of public health, arriving at conclusions that largely echo the clinical analysis. Both studies suggest that the fantasy of fully automated decision-making dissolves under close empirical scrutiny, replaced by a messier but more accurate account of humans and machines reasoning together.

The volume also confronts the hidden human labor on which algorithms depend. Cabitza and colleagues coin the neologism cybork, a fusion of cyborg and work, to emphasize that algorithms frequently rest on human effort. Bennett and colleagues explore the social and power dynamics that arise when humans clean data so that algorithms can operate more efficiently and more correctly. Their argument, aligned with Cabitza and colleagues, is that this labor should not be occluded as if the algorithms were completely autonomous. Instead, algorithmic processing should be understood as a form of collective networking and human-machine collaboration. This reframing carries real political weight: when the workers who make machine learning possible remain invisible, they also remain uncredited, unprotected, and excluded from conversations about how the resulting systems should be governed.

Entanglement, however, runs in both directions. Several essays trace the recursive feedback loops through which algorithms influence human behavior at the same time that humans shape algorithms. Lynch and colleagues demonstrate this dynamic at a microscale, examining how the introduction of robot tour guides into two Nevada museums helps to create the distinctive social spaces that visitors encounter. The robots are not merely props; they actively reorganize how people move, gather, and interact. Kalthoff and Link extend the analysis to the fabrication of humanoid robots designed to create social dynamics through their interactions with humans, citing the pioneering Kismet robot built by Cynthia Breazeal, an early attempt to engineer machines that could engage humans on an emotional register.

At the largest scale, Richard Groß argues that Large Language Models, through their stochastic procedures, are shaping social reality in their own image. A parallel argument has been advanced by Louise Amoore and her colleagues, who focus on the principal technical characteristics of LLMs, such as latency and tokenization, and contend that these systems create a world model that is rapidly propagating through the infrastructures of developed societies. The claim is technically grounded rather than metaphorical. Because LLMs mediate communication, search, and increasingly administrative processes, the statistical regularities embedded in their training and generation procedures can become the regularities of everyday social life, closing a loop in which human output trains the machines that then shape human output.

Beyond empirical case studies, the collection presses for a change in the frameworks, metaphors, and narratives that construct AI-human relations. Harry Halpin urges readers to think in terms of collective intelligence rather than artificial intelligence, a shift that relocates smartness from a single machine to a network of people and tools. Lewis and colleagues point toward indigenous knowledges as demonstrating what they call abundant intelligences that extend throughout the natural world. Hayles singles out this lesson as a general template for how communities can assume control over the kinds of AI that get created, the purposes those systems are designed to fulfill, and the human interests they can support and help to enact. The implication is that the trajectory of AI is not fixed by the technology itself but by the social arrangements that surround it.

Other contributors extend this critical reframing in complementary directions. Bennett and colleagues call for Responsible Artificial Intelligence, in which alignment with human values is treated as the primary design criterion rather than an afterthought. Bory and colleagues look critically at AI narratives themselves, arguing that visions with more modest ambitions, the so-called weak rather than strong versions of AI, hold more potential to lead to futures beneficial to humans. Nugent critiques the pervasive analogy between biological and technological evolution, showing that the analogy works only by suppressing crucial differences that reveal the complexities of biological processes. Together these essays suggest that the stories we tell about AI are not decoration; they actively determine which futures seem possible and which are dismissed before they are tried.

Perhaps the most technically provocative contribution comes from Fabian Offert and colleagues, whose essay on protein folding as predicted by AlphaFold and other Transformer technologies deconstructs the widespread belief that LLMs produce natural language. As the authors show, through tokenization, word embedding, and positional embedding, the linguistic qualities of language are taken apart, reconstructed as numbers in a high-dimensional vector space, and then reconstituted through stochastic procedures that differ radically from how natural language is used and understood by humans. What emerges is not language in any human sense but a new episteme, one that creates and reinforces an intuitive knowledge of sequence. That intuition about sequential structure, they suggest, is precisely what makes these architectures so powerful for predicting how chains of amino acids fold into functional proteins, a problem that had resisted solution for half a century.

The significance of this analysis extends well beyond molecular biology. Hayles notes that it resonates with her own criticism of the naturalness of machine language, and she argues that analyses like those of Offert and colleagues open new possibilities for literary criticism, cultural critique, and critical code studies. If the outputs of LLMs are the products of vector-space arithmetic and probabilistic sampling rather than meaning in the human sense, then humanists have a crucial role in explaining what these systems actually do, what they cannot do, and where their apparent fluency misleads the people who rely on it. The special issue thus positions the humanities not as a skeptical bystander to the AI revolution but as an essential participant in its technical and conceptual development.

The overall message of the collection is both a warning and an invitation. Hayles concludes that readers will find a plethora of diverse approaches, specific demonstrations, and game-changing ideas in these essays, all pointing to the ongoing necessity for critique and analysis by humanists, cultural critics, and others as powerful technologies develop at breakneck speed. We cannot afford to ignore AI, she writes, and the essays help chart paths that allow us to understand these systems, participate in their development, and have a say in how they are and will be used. In an era when headlines oscillate between utopian hype and existential dread, the volume offers a third way: careful, empirical attention to the actual entanglements of humans and machines, and a insistence that the shape of that entanglement remains a collective human choice.

Subject of Research: Human-AI entanglement and the mutual shaping of humans and algorithms in society

Article Title: Entangling humans and AI

Article References: Hayles, N. K. (2026). Entangling humans and AI. AI & SOCIETY. https://doi.org/10.1007/s00146-026-03387-x

Image Credits: AI Generated

DOI: 10.1007/s00146-026-03387-x

Keywords: artificial intelligence, human-AI collaboration, large language models, collective intelligence, algorithmic decision-making, data labor, responsible AI, protein folding, AlphaFold, AI narratives, posthumanism, AI & Society

The work, published in Nature by Chenyu Tang, Mayank Prakash Pandey and colleagues at the Université de Lorraine and CNRS in Nancy, France, with collaborators in Madrid, Chicago and Urbana-Champaign, addresses a stubborn bottleneck in computational biology. Standard molecular dynamics simulations propagate atoms step by step according to the laws of classical physics, and they do so faithfully. But the transitions that matter most, such as protein folding, allosteric signaling and membrane transport, are rare events. A simulation might churn through millions of time steps without ever witnessing the molecule cross from one stable state to another, because the probability of stumbling into the transition region by chance is vanishingly small.

Computational scientists have long responded with enhanced-sampling strategies, which apply artificial forces or cleverly designed biases to push molecules over energetic barriers. These methods have produced genuine insights, but they come with serious caveats. They are computationally demanding, and they typically require the researcher to choose, in advance, a set of collective variables, essentially a guess about which combination of atomic coordinates describes the reaction. If that guess is wrong or incomplete, the resulting free-energy landscape and apparent mechanism can be distorted. A recent study cited by the team put the problem bluntly in its title: convergence is not correctness, and enhanced-sampling methods can perform inconsistently as biological complexity increases.

Gen-COMPAS takes a different route. The framework couples a denoising diffusion probabilistic model, the same class of generative architecture that powers modern image synthesis, with a rigorous statistical concept from transition state theory known as the committor. The diffusion model is trained to produce structurally plausible intermediate conformations that sit between the two known end points of a transition. These generated structures are not accepted on faith. Instead, they are filtered using committor-based criteria, which identify the configurations from which a molecule is equally likely to fall forward to the product state or backward to the reactant state, the hallmark of a true transition state.

The committor is a powerful but demanding quantity. For any given molecular configuration, it answers a deceptively simple question: if the system were released from this point with random thermal velocities, what is the probability that it would commit to the product rather than the reactant? Configurations with a committor of one half define the transition state ensemble. Estimating this probability traditionally requires launching many short trajectories from each candidate configuration, which is exactly the kind of brute-force calculation that makes rare-event statistics so expensive. Gen-COMPAS turns this into an iterative loop: the generative model proposes candidate intermediates, short unbiased simulations from those intermediates test whether they genuinely belong to the transition region, and the results refine both the model and the ensemble.

The payoff is a dramatic compression of the required sampling. According to the authors, short unbiased simulations launched from the generated intermediates yield transition-region ensembles at aggregate sampling scales ranging from nanoseconds to submicroseconds, whereas conventional approaches would require orders of magnitude more simulation time to assemble equivalent statistics. In practical terms, the method replaces the hopeless task of waiting for a rare event to happen spontaneously with the tractable task of verifying that generated candidate structures are genuine members of the transition ensemble.

To demonstrate the framework, the team applied it to three systems of increasing size and biological complexity. The first was Trp-cage, a miniprotein of roughly twenty residues that has become a standard benchmark for folding studies because it folds quickly enough to be simulated directly. Decades of work, including landmark long-trajectory studies of fast-folding proteins, have mapped its folding landscape in detail, giving the team an independent yardstick against which to judge their results. Gen-COMPAS recovered committors, transition states and free-energy landscapes consistent with this established knowledge, starting only from the folded and unfolded end-point structures.

The second test system was the mitochondrial ADP/ATP carrier, a membrane transporter that shuttles the energy currency ATP across the inner mitochondrial membrane in exchange for ADP. This protein operates through a switching mechanism between conformational states that has been studied with structural biology and a variety of simulation techniques, but its full transition pathway remains challenging to sample because the carrier is embedded in a lipid membrane and undergoes large-scale domain movements. Gen-COMPAS reconstructed the transition pathway between the carrier’s known states without any predefined reaction coordinate, offering a mechanistic picture that previously would have demanded extensive biased simulations and careful, potentially contentious, choices of collective variables.

The third and most ambitious application was the nicotinic acetylcholine receptor, a pentameric ligand-gated ion channel that mediates synaptic transmission. Channel activation involves coordinated conformational changes across five protein subunits, and recent work has suggested that these subunit transitions occur asynchronously, priming the receptor for activation in a stepwise fashion. Simulating the full gating transition of such a large membrane protein is far beyond the reach of straightforward molecular dynamics. Gen-COMPAS nevertheless recovered the transition pathway and free-energy landscape from the known resting and activated structures alone, with no prior mechanistic knowledge encoded into the calculation.

The significance of the method lies not only in its speed but in what it does not require. No collective variables need to be chosen by hand, no reaction coordinates need to be guessed, and no prior hypothesis about the mechanism needs to be supplied. The generative model and the committor filter together let the data define the pathway. This removes one of the most persistent sources of bias in enhanced sampling, the researcher’s own assumptions about how the transition ought to proceed, and it opens the door to systems where no expert intuition exists to guide variable selection.

The method also arrives amid a broader convergence of machine learning and molecular simulation. Generative models have been used to emulate equilibrium protein ensembles, to sample molecular dynamics trajectories directly, and to guide transition path sampling in systems ranging from lipid flip-flop to membrane nanoporation. Gen-COMPAS distinguishes itself by combining generative proposal with a physically rigorous acceptance criterion rooted in transition rate theory, ensuring that the structures it produces are not merely plausible but statistically meaningful members of the true transition ensemble. The authors acknowledge prior machine-guided path-sampling work and committor-learning approaches as foundations for this synthesis.

Accessibility may prove as important as performance. Gen-COMPAS is released as an installable open-source package on GitHub, complete with a unified command-line interface, ready-to-use configuration files, molecular-simulation inputs for representative systems, and a graphical user interface that guides users through the available options. Input files for all systems analyzed in the paper are publicly available, and source data accompany the publication. The team also gratefully acknowledges reference molecular dynamics trajectories provided by D. E. Shaw Research, whose specialized hardware has long set the benchmark for long unbiased simulations that methods like this one can now be validated against.

For the field, the implications extend across drug discovery, basic biophysics and beyond. Ion channels and membrane transporters are among the most heavily targeted protein families in pharmacology, and understanding the exact pathway by which they switch states can reveal where small molecules might bind to stabilize one state or another. Allosteric transitions in signaling proteins, another class of rare events, could similarly be mapped without the guesswork that has historically limited such studies. If the framework generalizes as its demonstration systems suggest, the computational microscope that biophysicists have been building for half a century may finally be able to zoom in on the fleeting moments that matter most, the instants in which a molecule decides what it will become next.

Subject of Research: Generative committor-guided path sampling of biomolecular conformational transitions

Article Title: Breaking timescales with generative sampling of conformational transitions

Article References: Tang, C., Pandey, M. P., Chen, C. G., Megías, A., Dehez, F., & Chipot, C. (2026). Breaking timescales with generative sampling of conformational transitions. Nature, 657(8133), 944-949. https://doi.org/10.1038/s41586-026-11025-1

Image Credits: AI Generated

DOI: 10.1038/s41586-026-11025-1

Keywords: Gen-COMPAS, molecular dynamics, transition path sampling, denoising diffusion model, committor, protein folding, enhanced sampling, ion channel, membrane transporter, free-energy landscape, machine learning, computational biophysics

The importance of beta-barrel proteins to mitochondrial biology is difficult to overstate. TOM40 forms the primary channel through which nearly all mitochondrial proteins enter the organelle, while the VDACs regulate the flow of metabolites such as ATP and ions across the outer membrane. SAM50, the central pore-forming subunit of the SAM complex, belongs to the Omp85 family of bacterial and organellar proteins, and it carries a distinctive structural feature called a lateral gate. This gate is a region of the barrel that can open sideways, allowing a nascent beta-barrel substrate to partially release from the SAM50 lumen and begin folding into its own barrel within the membrane. Decades of work on bacterial and mitochondrial assembly machineries have established that barrel folding proceeds hand in hand with conformational changes at these lateral gates, making the gate a dynamic control point rather than a static seam.

In fungi, where the SAM complex consists of Sam50 plus the peripheral subunits Sam35 and Sam37, high-resolution structural studies have established what researchers call a beta-barrel-switching model. In this model, a second beta-barrel protein binds dynamically at a site immediately adjacent to the lateral gate of Sam50, effectively handing off the growing substrate from one barrel to another as folding proceeds. The fungal structures showed this switching site occupied by barrel proteins caught in the act of assembly, providing a snapshot of the folding pathway in progress. But a critical question has lingered: do mammalian SAM complexes, which use entirely different peripheral subunits, work the same way? In humans, the proteins METAXIN1 and METAXIN2 replace Sam37 and Sam35 respectively, and whether the human machine employs an equivalent switching mechanism had remained unresolved.

The new study answers that question with striking structural detail. The research team, including co-first authors Ling Yan, Zeyuan Guan and Qiang Wang, determined cryo-EM structures of two distinct forms of the human SAM complex. The first is a monomeric complex containing a single SAM50 molecule paired with one METAXIN2, forming a SAM50-METAXIN2 core. The second is a dimeric complex containing two SAM50 molecules, again with one METAXIN2 associated. The atomic coordinates and cryo-EM maps for both structures have been deposited in public databases, with the monomeric complex archived under PDB code 25WG and EMDB entry EMD-80424, and the dimeric complex under PDB 25WF and EMD-80423, allowing researchers worldwide to examine the molecular details directly.

The dimeric structure is the showstopper. Rather than simply sitting side by side, the two SAM50 barrels adopt what the authors describe as an interlocked conformation: each SAM50 molecule extends beta-strands into the barrel lumen of its partner. This molecular intertwining means that one SAM50 barrel physically occupies the beta-barrel-switching site adjacent to the lateral gate of the other SAM50. In other words, the dimer captures the human SAM complex in a configuration that directly mirrors the fungal beta-barrel-switching model, but with the remarkable twist that the second barrel at the switching site is another copy of SAM50 itself. This suggests that in human mitochondria, a second SAM50 molecule may serve as a placeholder or scaffold at the switching site, potentially priming the machine for substrate engagement or reflecting a regulatory state in the assembly cycle.

To test whether this dimeric arrangement is functionally meaningful rather than a crystallographic artifact of sample preparation, the team turned to biochemical and cell-based approaches. Crosslinking experiments, which chemically freeze transient protein-protein contacts so they can be detected, provided independent evidence for the close apposition of the two SAM50 molecules in living systems. More tellingly, the researchers employed fluorescence lifetime imaging microscopy combined with Foerster resonance energy transfer, a technique known as FLIM-FRET that reports on nanometer-scale distances between fluorescently labeled proteins inside cells. These analyses revealed that METAXIN1, the mammalian counterpart of fungal Sam37, alters the conformation of SAM50 and changes the relative arrangement of the two SAM50 molecules in the dimeric complex.

This METAXIN1-dependent remodeling is a key finding because it demonstrates that the dimeric SAM assembly is not a rigid, static structure but a dynamic platform whose geometry is actively tuned by peripheral subunits. In the fungal system, Sam37 plays roles in substrate release from the SAM complex, and the implication is that METAXIN1 may perform an analogous regulatory function in human cells, repositioning the SAM50 barrels to facilitate the progression of substrate folding and release. The structural and functional data together suggest a working model in which the human SAM complex cycles between monomeric and dimeric states, with METAXIN1 and METAXIN2 orchestrating conformational transitions that drive beta-barrel substrates through the folding pathway.

The evolutionary dimension of this work is equally compelling. The fact that the beta-barrel-switching principle is conserved from fungi to humans, despite the complete replacement of the peripheral subunits Sam35 and Sam37 with METAXIN2 and METAXIN1, indicates that the core mechanism of barrel assembly is ancient and tightly constrained. The Omp85 family spans bacteria, chloroplasts and mitochondria, and lateral-gate dynamics have been documented across all these systems. What the new structures add is the demonstration that the switching site itself, the physical location where a second barrel docks next to the gate, is a conserved feature of the mitochondrial machinery, even as the molecular players that regulate it have been swapped out during evolution. This kind of architectural conservation with component turnover is a recurring theme in mitochondrial biology and underscores the deep common ancestry of the organelle’s protein import systems.

Why should this matter beyond structural biology? Defects in mitochondrial protein import and outer membrane assembly are increasingly linked to human disease, and the VDACs and TOM40 in particular have been implicated in neurodegeneration, cancer metabolism and mitochondrial dysfunction more broadly. By defining the conformational states of the human SAM complex and identifying METAXIN1 as a remodeling factor, the study establishes a framework that other researchers can use to probe how barrel assembly goes wrong in disease contexts, and potentially how it could be modulated. The structures also provide concrete molecular targets for future studies of METAXIN function, which until now have been largely mysterious in mechanistic terms.

The work also showcases the power of modern cryo-electron microscopy to resolve small, dynamic membrane protein assemblies that were once considered beyond reach. Capturing both the monomeric and interlocked dimeric states of the human SAM complex, complete with the beta-strand exchange between barrels, required both technical skill in sample preparation and data collection, acknowledged by the authors to the cryo-EM facilities at the University of Science and Technology of China and Huazhong Agricultural University, and sophisticated computational reconstruction. As cryo-EM continues to push toward smaller and more heterogeneous complexes, studies like this one are filling in the remaining gaps in our understanding of how cells build their most essential molecular machines. For the mitochondrial outer membrane, the picture is now coming into focus: a conserved barrel-switching engine, remodeled by mammalian-specific regulators, assembling the channels that keep cellular power plants running.

Subject of Research: Cryo-EM structures of the human mitochondrial SAM complex involved in beta-barrel protein assembly

Article Title: Cryo-EM structures of monomeric and dimeric human SAM complexes

Article References: Yan, L., Guan, Z., Wang, Q., Wu, Y., Qi, L., Huang, C., Zhang, J., Hong, S., Zhuang, J., Huang, R., Zhang, L., Duan, R., Liu, Y., Liu, Z., & Yin, P. (2026). Cryo-EM structures of monomeric and dimeric human SAM complexes. Cell Research. https://doi.org/10.1038/s41422-026-01305-w

Image Credits: AI Generated

DOI: 10.1038/s41422-026-01305-w

Keywords: cryo-EM, mitochondria, SAM complex, SAM50, beta-barrel proteins, METAXIN1, METAXIN2, protein folding, outer mitochondrial membrane, TOM40, VDAC, structural biology

The study’s central engine was summary-data-based Mendelian randomization, or SMR, a statistical technique that exploits naturally occurring genetic variation to probe whether molecular traits influence disease risk. Because genetic variants are randomly allocated at conception, much like coins flipped by nature, they are largely immune to the confounding and reverse causation that plague conventional observational studies. In an SMR analysis, variants that alter, say, the methylation level of a DNA site can be used as instruments to ask whether that methylation change has a downstream causal effect on disease. The method requires that the same variant be associated with both the molecular trait and the outcome, and the strength of the inferred relationship depends on the biology linking the two.

What made this study ambitious was the stacking of three molecular layers on top of a genome-wide association study (GWAS) of major depressive disorder. The researchers merged the MDD GWAS summary statistics with blood-derived cis-methylation quantitative trait loci (mQTLs), cis-expression quantitative trait loci (eQTLs), and cis-protein quantitative trait loci (pQTLs) datasets. Each layer captures a different rung on the ladder from DNA to function: methylation marks influence how genes are regulated, eQTLs reveal how those regulatory differences change gene expression, and pQTLs show how genetic variation alters the abundance of proteins circulating in the blood. By screening candidate signals at every rung, the team could trace a provisional causal chain from epigenetic regulation through gene activity to protein level and finally to disease risk.

The initial screening in the discovery cohort yielded a substantial haul. Eighty-six methylation sites, fifteen genes, and three proteins passed the prespecified SMR thresholds, suggesting that protein folding-related biology was repeatedly intersecting with depression-associated genetic signals. But genomic screening is notoriously prone to false positives, so the investigators applied a false discovery rate (FDR) correction to rein in spurious findings. After this stricter accounting, eleven methylation sites mapping to five genes remained statistically significant at the methylation layer. The expression and protein-level signals, by contrast, did not survive multiple-testing correction and were classified as nominal or suggestive. This hierarchy of evidence is itself informative: it points to methylation, the layer closest to gene regulation, as the tier where protein folding-related genetics and depression risk most clearly converge.

Colocalization analysis provided a second, independent filter. When two traits appear to share a genetic association, there is always a possibility that the signal is actually produced by two different variants sitting near each other on the same chromosome, a phenomenon rooted in linkage disequilibrium. Colocalization methods test whether the two association signals are driven by the same causal variant. In this study, colocalization supported shared genetic signals for ten methylation sites, four expression-associated genes, and two proteins. Of those, four methylation signals, two expression signals, and one protein signal showed strong colocalization evidence, with posterior probabilities exceeding 0.8, a benchmark widely regarded as compelling. Signals that survive both SMR and colocalization are far less likely to be artifacts of chromosomal proximity.

Cross-omic integration then allowed the researchers to stitch the layers together. The analyses provided genetically supported evidence for potential regulatory relationships between FNIP2 methylation-related signals and the gene’s expression, and between PDIA3 expression and its protein abundance. In plain terms, the data suggest a plausible chain of causation in which inherited variation alters chemical tags on the FNIP2 gene, which in turn shifts how actively the gene is expressed; separately, variants affecting PDIA3 expression appear to propagate upward to change the amount of PDIA3 protein detectable in blood. Both molecular traits also showed exploratory associations with MDD risk across their corresponding omic layers, hinting at complete, if provisional, chains from variant to molecule to disorder.

The two prioritized genes are biologically intriguing in their own right. PDIA3 encodes a protein disulfide isomerase resident in the endoplasmic reticulum, where it catalyzes the reshuffling of disulfide bonds that allow newly made proteins to assume their correct three-dimensional shapes. FNIP2 interacts with folliculin and participates in AMP-activated protein kinase (AMPK) signaling, a cellular energy-sensing pathway with documented ties to stress responses. Neither gene is a household name in depression research, which is precisely why a systematic, hypothesis-free approach was needed to surface them. Their emergence from an unbiased screen suggests that depression genetics may be whispering about cellular stress biology that conventional candidate-gene studies have overlooked.

Functional enrichment analyses reinforced that interpretation. Gene Ontology and KEGG pathway analyses of the candidate genes implicated proteostasis-related modules, including endoplasmic reticulum stress, the unfolded protein response, chaperone-mediated folding, protein processing, quality control, and antigen presentation. The unfolded protein response is the cell’s emergency program when misfolded proteins accumulate in the endoplasmic reticulum, and chronic activation of this stress pathway has been observed in animal models of depression, including those using chronic unpredictable mild stress. The enrichment of antigen presentation and major histocompatibility complex-related terms also dovetails with a growing literature connecting immune dysregulation to mood disorders, suggesting that protein folding quality control and inflammation may be intertwined strands of the same biological rope.

The team also sought external support through additional analyses, though these remained exploratory. Replication attempts in the FinnGen depression dataset, cross-disorder checks against bipolar disorder, and surveys of brain-region expression data drawn from the Gene Expression Omnibus provided directionally consistent but not definitive signals. The authors are appropriately measured in their conclusions: the findings are genetically supported and hypothesis-generating, and they explicitly call for independent replication and functional validation before PDIA3 or FNIP2 can be considered established depression genes. Genetic instruments indicate association with disease risk through molecular traits; they do not, by themselves, prove how the genes act in neurons or glia.

Even with those caveats, the study’s design offers a template for the next generation of psychiatric genetics. Rather than asking which single variant raises disease risk, multi-omics Mendelian randomization asks which molecular mechanisms inheritable variation plausibly perturbs, and then interrogates those mechanisms layer by layer. The prioritization of protein folding and proteostasis pathways in major depressive disorder reframes depression not merely as a disorder of neurotransmitters but as a condition in which cellular stress, protein quality control and immune signaling may help set the threshold at which adversity tips into illness. If PDIA3 and FNIP2 hold up under replication and laboratory scrutiny, they could point toward biomarkers measurable in blood and, ultimately, toward therapeutic strategies that shore up the cell’s faltering protein-folding machinery in the most vulnerable patients.

One methodological detail worth noting is how the authors guarded against a known weakness of SMR: horizontal pleiotropy, in which an instrument variant influences the disease through a pathway unrelated to the molecular trait under study. The HEIDI test addresses this by examining whether the association between the instrument and the outcome is consistent across many variants scattered across the locus. A genuine causal effect should show a uniform signal, whereas linkage-driven artifacts tend to concentrate among variants closest to the probe. By requiring HEIDI testing alongside colocalization, the analysis applied two complementary safeguards against the same class of false positive, which strengthens confidence in the methylation-layer findings that survived both filters.

The choice of blood as the tissue source for all three molecular layers also deserves consideration. Blood is far easier to sample than brain tissue, which is why most large-scale QTL reference panels are built from it, and peripheral methylation and protein signals can serve as accessible biomarkers. Yet depression is a disorder of the brain, and regulatory biology in blood does not always mirror that in neural tissue. The exploratory surveys of brain-region expression data, including the anterior cingulate cortex, represent an early attempt to bridge this gap, and the authors themselves flag these analyses as preliminary rather than confirmatory.

The reporting of the study followed STROBE-MR guidelines, a checklist designed to improve transparency in Mendelian randomization research by requiring explicit documentation of instrument selection, sensitivity analyses, and potential pleiotropy. Such standardized reporting matters because genetic instruments can fail in subtle ways, and readers need enough detail to judge whether assumptions hold. The open-access publication also means that the supplementary tables, which contain the full sets of screened methylation sites, genes, and proteins, are available to any laboratory wishing to reanalyze the signals or design follow-up experiments targeting PDIA3 or FNIP2 in cellular models of stress.

Subject of Research: Genetically supported associations between protein folding-related genes and major depressive disorder identified through multi-omics Mendelian randomization.

Article Title: Unraveling the role of protein folding-related genes in depression through multi-omics mendelian randomization

Article References: Wang, J., He, S., Cui, J., & Li, H. (2026). Unraveling the role of protein folding-related genes in depression through multi-omics mendelian randomization. Annals of General Psychiatry. https://doi.org/10.1186/s12991-026-00697-8

Image Credits: AI Generated

DOI: 10.1186/s12991-026-00697-8

Keywords: depression, major depressive disorder, protein folding, Mendelian randomization, multi-omics, PDIA3, FNIP2, DNA methylation, colocalization, unfolded protein response, genetics, psychiatry