The 100 most-cited articles in Nature, 2024

Abstract The introduction of AlphaFold 2 1 has spurred a revolution in modelling the structure of proteins and their interactions, enabling a huge range of applications in protein modelling and design 2–6 . Here we describe our AlphaFold 3 model with a substantially updated diffusion-based architecture that is capable of predicting the joint structure of complexes including proteins, nucleic acids, small molecules, ions and modified residues. The new AlphaFold model demonstrates substantially improved accuracy over many previous specialized tools: far greater accuracy for protein–ligand interactions compared with state-of-the-art docking tools, much higher accuracy for protein–nucleic acid interactions compared with nucleic-acid-specific predictors and substantially higher antibody–antigen prediction accuracy compared with AlphaFold-Multimer v.2.3 7,8 . Together, these results show that high-accuracy modelling across biomolecular space is possible within a single unified deep-learning framework.

Abstract Large language model (LLM) systems, such as ChatGPT 1 or Gemini 2 , can show impressive reasoning and question-answering capabilities but often ‘hallucinate’ false outputs and unsubstantiated answers 3,4 . Answering unreliably or without the necessary information prevents adoption in diverse fields, with problems including fabrication of legal precedents 5 or untrue facts in news articles 6 and even posing a risk to human life in medical domains such as radiology 7 . Encouraging truthfulness through supervision or reinforcement has been only partially successful 8 . Researchers need a general method for detecting hallucinations in LLMs that works even with new and unseen questions to which humans might not know the answer. Here we develop new methods grounded in statistics, proposing entropy-based uncertainty estimators for LLMs to detect a subset of hallucinations—confabulations—which are arbitrary and incorrect generations. Our method addresses the fact that one idea can be expressed in many ways by computing uncertainty at the level of meaning rather than specific…

Abstract Digital pathology poses unique computational challenges, as a standard gigapixel slide may comprise tens of thousands of image tiles 1–3 . Prior models have often resorted to subsampling a small portion of tiles for each slide, thus missing the important slide-level context 4 . Here we present Prov-GigaPath, a whole-slide pathology foundation model pretrained on 1.3 billion 256 × 256 pathology image tiles in 171,189 whole slides from Providence, a large US health network comprising 28 cancer centres. The slides originated from more than 30,000 patients covering 31 major tissue types. To pretrain Prov-GigaPath, we propose GigaPath, a novel vision transformer architecture for pretraining gigapixel pathology slides. To scale GigaPath for slide-level learning with tens of thousands of image tiles, GigaPath adapts the newly developed LongNet 5 method to digital pathology. To evaluate Prov-GigaPath, we construct a digital pathology benchmark comprising 9 cancer subtyping tasks and 17 pathomics tasks, using both Providence and TCGA data 6 . With large-scale pretraining and…

Abstract Comprehensively mapping the genetic basis of human disease across diverse individuals is a long-standing goal for the field of human genetics 1–4 . The All of Us Research Program is a longitudinal cohort study aiming to enrol a diverse group of at least one million individuals across the USA to accelerate biomedical research and improve human health 5,6 . Here we describe the programme’s genomics data release of 245,388 clinical-grade genome sequences. This resource is unique in its diversity as 77% of participants are from communities that are historically under-represented in biomedical research and 46% are individuals from under-represented racial and ethnic minorities. All of Us identified more than 1 billion genetic variants, including more than 275 million previously unreported genetic variants, more than 3.9 million of which had coding consequences. Leveraging linkage between genomic data and the longitudinal electronic health record, we evaluated 3,724 genetic variants associated with 117 diseases and found high replication rates across both participants of European…

) demonstrated high performance across a variety of language tasks. ChatGPT introduced such language models to the public. It is now clear that generative artificial intelligence (AI) such as large language models (LLMs) is here to stay and will substantially change the ecosystem of online text and images. Here we consider what may happen to GPT-{n} once LLMs contribute much of the text found online. We find that indiscriminate use of model-generated content in training causes irreversible defects in the resulting models, in which tails of the original content distribution disappear. We refer to this effect as 'model collapse' and show that it can occur in LLMs as well as in variational autoencoders (VAEs) and Gaussian mixture models (GMMs). We build theoretical intuition behind the phenomenon and portray its ubiquity among all learned generative models. We demonstrate that it must be taken seriously if we are to sustain the benefits of training from large-scale data scraped from the web. Indeed, the value of data collected about genuine human interactions with systems will be…

Quantum error correction [1, 2, 3, 4] provides a path to reach practical quantum computing by combining multiple physical qubits into a logical qubit, where the logical error rate is suppressed exponentially as more qubits are added. However, this exponential suppression only occurs if the physical error rate is below a critical threshold. Here, we present two below-threshold surface code memories on our newest generation of superconducting processors, Willow: a distance-7 code, and a distance-5 code integrated with a real-time decoder. The logical error rate of our larger quantum memory is suppressed by a factor of $$\varLambda \mathrm{=2.14}\pm 0.02$$ when increasing the code distance by two, culminating in a 101-qubit distance-7 code with 0.143% ± 0.003% error per cycle of error correction. This logical memory is also beyond break-even, exceeding its best physical qubit’s lifetime by a factor of $$2.4\pm 0.3$$ . Our system maintains below-threshold performance when decoding in real time, achieving an average decoder latency of 63 μs at distance-5 up to a million cycles, with a…

The uptake of carbon dioxide (CO2) by terrestrial ecosystems is critical for moderating climate change1. To provide a ground-based long-term assessment of the contribution of forests to terrestrial CO2 uptake, we synthesized in situ forest data from boreal, temperate and tropical biomes spanning three decades. We found that the carbon sink in global forests was steady, at 3.6 ± 0.4 Pg C yr−1 in the 1990s and 2000s, and 3.5 ± 0.4 Pg C yr−1 in the 2010s. Despite this global stability, our analysis revealed some major biome-level changes. Carbon sinks have increased in temperate (+30 ± 5%) and tropical regrowth (+29 ± 8%) forests owing to increases in forest area, but they decreased in boreal (−36 ± 6%) and tropical intact (−31 ± 7%) forests, as a result of intensified disturbances and losses in intact forest area, respectively. Mass-balance studies indicate that the global land carbon sink has increased2, implying an increase in the non-forest-land carbon sink. The global forest sink is equivalent to almost half of fossil-fuel emissions (7.8 ± 0.4 Pg C yr−1 in 1990–2019). However,…

Abstract Groundwater resources are vital to ecosystems and livelihoods. Excessive groundwater withdrawals can cause groundwater levels to decline 1–10 , resulting in seawater intrusion 11 , land subsidence 12,13 , streamflow depletion 14–16 and wells running dry 17 . However, the global pace and prevalence of local groundwater declines are poorly constrained, because in situ groundwater levels have not been synthesized at the global scale. Here we analyse in situ groundwater-level trends for 170,000 monitoring wells and 1,693 aquifer systems in countries that encompass approximately 75% of global groundwater withdrawals 18 . We show that rapid groundwater-level declines (>0.5 m year −1 ) are widespread in the twenty-first century, especially in dry regions with extensive croplands. Critically, we also show that groundwater-level declines have accelerated over the past four decades in 30% of the world’s regional aquifers. This widespread acceleration in groundwater-level deepening highlights an urgent need for more effective measures to address groundwater depletion. Our analysis…

Abstract Lifted Kramers spin degeneracy (LKSD) has been among the central topics of condensed-matter physics since the dawn of the band theory of solids 1,2 . It underpins established practical applications as well as current frontier research, ranging from magnetic-memory technology 3–7 to topological quantum matter 8–14 . Traditionally, LKSD has been considered to originate from two possible internal symmetry-breaking mechanisms. The first refers to time-reversal symmetry breaking by magnetization of ferromagnets and tends to be strong because of the non-relativistic exchange origin 15 . The second applies to crystals with broken inversion symmetry and tends to be comparatively weaker, as it originates from the relativistic spin–orbit coupling (SOC) 16–19 . A recent theory work based on spin-symmetry classification has identified an unconventional magnetic phase, dubbed altermagnetic 20,21 , that allows for LKSD without net magnetization and inversion-symmetry breaking. Here we provide the confirmation using photoemission spectroscopy and ab initio calculations. We identify two…

Interpreting electron cryo-microscopy (cryo-EM) maps with atomic models requires high levels of expertise and labour-intensive manual intervention. We present ModelAngelo, a machine-learning approach for automated atomic model building in cryo-EM maps. By combining information from the cryo-EM map with information from protein sequence and structure in a single graph neural network, ModelAngelo builds atomic models for proteins that are of similar quality as those generated by human experts. For nucleotides, ModelAngelo builds backbones with similar accuracy as humans. By using its predicted amino acid probabilities for each residue in hidden Markov model sequence searches, ModelAngelo outperforms human experts in the identification of proteins with unknown sequences. ModelAngelo will thus remove bottlenecks and increase objectivity in cryo-EM structure determination.

Abstract The possibility that the Amazon forest system could soon reach a tipping point, inducing large-scale collapse, has raised global concern 1–3 . For 65 million years, Amazonian forests remained relatively resilient to climatic variability. Now, the region is increasingly exposed to unprecedented stress from warming temperatures, extreme droughts, deforestation and fires, even in central and remote parts of the system 1 . Long existing feedbacks between the forest and environmental conditions are being replaced by novel feedbacks that modify ecosystem resilience, increasing the risk of critical transition. Here we analyse existing evidence for five major drivers of water stress on Amazonian forests, as well as potential critical thresholds of those drivers that, if crossed, could trigger local, regional or even biome-wide forest collapse. By combining spatial information on various disturbances, we estimate that by 2050, 10% to 47% of Amazonian forests will be exposed to compounding disturbances that may trigger unexpected ecosystem transitions and potentially exacerbate…

Abstract Type 2 diabetes (T2D) is a heterogeneous disease that develops through diverse pathophysiological processes 1,2 and molecular mechanisms that are often specific to cell type 3,4 . Here, to characterize the genetic contribution to these processes across ancestry groups, we aggregate genome-wide association study data from 2,535,601 individuals (39.7% not of European ancestry), including 428,452 cases of T2D. We identify 1,289 independent association signals at genome-wide significance ( P < 5 × 10 −8 ) that map to 611 loci, of which 145 loci are, to our knowledge, previously unreported. We define eight non-overlapping clusters of T2D signals that are characterized by distinct profiles of cardiometabolic trait associations. These clusters are differentially enriched for cell-type-specific regions of open chromatin, including pancreatic islets, adipocytes, endothelial cells and enteroendocrine cells. We build cluster-specific partitioned polygenic scores 5 in a further 279,552 individuals of diverse ancestry, including 30,288 cases of T2D, and test their association with…

Abstract The Warburg effect is a hallmark of cancer that refers to the preference of cancer cells to metabolize glucose anaerobically rather than aerobically 1,2 . This results in substantial accumulation of lacate, the end product of anaerobic glycolysis, in cancer cells 3 . However, how cancer metabolism affects chemotherapy response and DNA repair in general remains incompletely understood. Here we report that lactate-driven lactylation of NBS1 promotes homologous recombination (HR)-mediated DNA repair. Lactylation of NBS1 at lysine 388 (K388) is essential for MRE11–RAD50–NBS1 (MRN) complex formation and the accumulation of HR repair proteins at the sites of DNA double-strand breaks. Furthermore, we identify TIP60 as the NBS1 lysine lactyltransferase and the ‘writer’ of NBS1 K388 lactylation, and HDAC3 as the NBS1 de-lactylase. High levels of NBS1 K388 lactylation predict poor patient outcome of neoadjuvant chemotherapy, and lactate reduction using either genetic depletion of lactate dehydrogenase A (LDHA) or stiripentol, a lactate dehydrogenase A inhibitor used clinically for…

Abstract Connections between neurons can be mapped by acquiring and analysing electron microscopic brain images. In recent years, this approach has been applied to chunks of brains to reconstruct local connectivity maps that are highly informative 1–6 , but nevertheless inadequate for understanding brain function more globally. Here we present a neuronal wiring diagram of a whole brain containing 5 × 10 7 chemical synapses 7 between 139,255 neurons reconstructed from an adult female Drosophila melanogaster 8,9 . The resource also incorporates annotations of cell classes and types, nerves, hemilineages and predictions of neurotransmitter identities 10–12 . Data products are available for download, programmatic access and interactive browsing and have been made interoperable with other fly data resources. We derive a projectome—a map of projections between regions—from the connectome and report on tracing of synaptic pathways and the analysis of information flow from inputs (sensory and ascending neurons) to outputs (motor, endocrine and descending neurons) across both hemispheres and…

. However, the relationship between lipid metabolism in glia and Alzheimer's disease pathology remains poorly understood. Through single-nucleus RNA sequencing of brain tissue in Alzheimer's disease, we have identified a microglial state defined by the expression of the lipid droplet-associated enzyme ACSL1 with ACSL1-positive microglia being most abundant in patients with Alzheimer's disease having the APOE4/4 genotype. In human induced pluripotent stem cell-derived microglia, fibrillar Aβ induces ACSL1 expression, triglyceride synthesis and lipid droplet accumulation in an APOE-dependent manner. Additionally, conditioned media from lipid droplet-containing microglia lead to Tau phosphorylation and neurotoxicity in an APOE-dependent manner. Our findings suggest a link between genetic risk factors for Alzheimer's disease with microglial lipid droplet accumulation and neurotoxic microglia-derived factors, potentially providing therapeutic strategies for Alzheimer's disease.

Highly pathogenic avian influenza (HPAI) H5N1 clade 2.3.4.4b virus has caused the death of millions of domestic birds and thousands of wild birds in the U.S. since January, 2022 1–4 Throughout this outbreak, spillovers to mammals have been frequently documented 5–12 . We report spillover of HPAI H5N1 virus in dairy cattle herds across several states in the U.S. The affected cows displayed clinical signs encompassing decreased feed intake, altered fecal consistency, respiratory distress, and decreased milk production with abnormal milk. Infectious virus and viral RNA were consistently detected in milk from affected cows. Viral distribution in tissues via immunohistochemistry and in situ hybridization revealed a distinct tropism of the virus for the epithelial cells lining the alveoli of the mammary gland in cows. Whole viral genome sequences recovered from dairy cows, birds, domestic cats, and a raccoon from affected farms indicated multidirectional interspecies transmissions. Epidemiologic and genomic data revealed efficient cow-to-cow transmission after apparently healthy cows from…

Abstract Computational pathology 1,2 has witnessed considerable progress in the development of both task-specific predictive models and task-agnostic self-supervised vision encoders 3,4 . However, despite the explosive growth of generative artificial intelligence (AI), there have been few studies on building general-purpose multimodal AI assistants and copilots 5 tailored to pathology. Here we present PathChat, a vision-language generalist AI assistant for human pathology. We built PathChat by adapting a foundational vision encoder for pathology, combining it with a pretrained large language model and fine-tuning the whole system on over 456,000 diverse visual-language instructions consisting of 999,202 question and answer turns. We compare PathChat with several multimodal vision-language AI assistants and GPT-4V, which powers the commercially available multimodal general-purpose AI assistant ChatGPT-4 (ref. 6 ). PathChat achieved state-of-the-art performance on multiple-choice diagnostic questions from cases with diverse tissue origins and disease models. Furthermore, using…

Abstract General circulation models (GCMs) are the foundation of weather and climate prediction 1,2 . GCMs are physics-based simulators that combine a numerical solver for large-scale dynamics with tuned representations for small-scale processes such as cloud formation. Recently, machine-learning models trained on reanalysis data have achieved comparable or better skill than GCMs for deterministic weather forecasting 3,4 . However, these models have not demonstrated improved ensemble forecasts, or shown sufficient stability for long-term weather and climate simulations. Here we present a GCM that combines a differentiable solver for atmospheric dynamics with machine-learning components and show that it can generate forecasts of deterministic weather, ensemble weather and climate on par with the best machine-learning and physics-based methods. NeuralGCM is competitive with machine-learning models for one- to ten-day forecasts, and with the European Centre for Medium-Range Weather Forecasts ensemble prediction for one- to fifteen-day forecasts. With prescribed sea surface temperature,…

Abstract The accumulation of physical errors 1–3 prevents the execution of large-scale algorithms in current quantum computers. Quantum error correction 4 promises a solution by encoding k logical qubits onto a larger number n of physical qubits, such that the physical errors are suppressed enough to allow running a desired computation with tolerable fidelity. Quantum error correction becomes practically realizable once the physical error rate is below a threshold value that depends on the choice of quantum code, syndrome measurement circuit and decoding algorithm 5 . We present an end-to-end quantum error correction protocol that implements fault-tolerant memory on the basis of a family of low-density parity-check codes 6 . Our approach achieves an error threshold of 0.7% for the standard circuit-based noise model, on par with the surface code 7–10 that for 20 years was the leading code in terms of error threshold. The syndrome measurement cycle for a length- n code in our family requires n ancillary qubits and a depth-8 circuit with CNOT gates, qubit initializations and…

Abstract Fusobacterium nucleatum ( Fn ), a bacterium present in the human oral cavity and rarely found in the lower gastrointestinal tract of healthy individuals 1 , is enriched in human colorectal cancer (CRC) tumours 2–5 . High intratumoural Fn loads are associated with recurrence, metastases and poorer patient prognosis 5–8 . Here, to delineate Fn genetic factors facilitating tumour colonization, we generated closed genomes for 135 Fn strains; 80 oral strains from individuals without cancer and 55 unique cancer strains cultured from tumours from 51 patients with CRC. Pangenomic analyses identified 483 CRC-enriched genetic factors. Tumour-isolated strains predominantly belong to Fn subspecies animalis ( Fna ). However, genomic analyses reveal that Fna , considered a single subspecies, is instead composed of two distinct clades ( Fna C1 and Fna C2). Of these, only Fna C2 dominates the CRC tumour niche. Inter- Fna analyses identified 195 Fna C2-associated genetic factors consistent with increased metabolic potential and colonization of the gastrointestinal tract. In support of this,…

Abstract Angiosperms are the cornerstone of most terrestrial ecosystems and human livelihoods 1,2 . A robust understanding of angiosperm evolution is required to explain their rise to ecological dominance. So far, the angiosperm tree of life has been determined primarily by means of analyses of the plastid genome 3,4 . Many studies have drawn on this foundational work, such as classification and first insights into angiosperm diversification since their Mesozoic origins 5–7 . However, the limited and biased sampling of both taxa and genomes undermines confidence in the tree and its implications. Here, we build the tree of life for almost 8,000 (about 60%) angiosperm genera using a standardized set of 353 nuclear genes 8 . This 15-fold increase in genus-level sampling relative to comparable nuclear studies 9 provides a critical test of earlier results and brings notable change to key groups, especially in rosids, while substantiating many previously predicted relationships. Scaling this tree to time using 200 fossils, we discovered that early angiosperm evolution was characterized by…

Abstract Weather forecasts are fundamentally uncertain, so predicting the range of probable weather scenarios is crucial for important decisions, from warning the public about hazardous weather to planning renewable energy use. Traditionally, weather forecasts have been based on numerical weather prediction (NWP) 1 , which relies on physics-based simulations of the atmosphere. Recent advances in machine learning (ML)-based weather prediction (MLWP) have produced ML-based models with less forecast error than single NWP simulations 2,3 . However, these advances have focused primarily on single, deterministic forecasts that fail to represent uncertainty and estimate risk. Overall, MLWP has remained less accurate and reliable than state-of-the-art NWP ensemble forecasts. Here we introduce GenCast, a probabilistic weather model with greater skill and speed than the top operational medium-range weather forecast in the world, ENS, the ensemble forecast of the European Centre for Medium-Range Weather Forecasts 4 . GenCast is an ML weather prediction method, trained on decades of reanalysis…

Abstract The fruit fly Drosophila melanogaster has emerged as a key model organism in neuroscience, in large part due to the concentration of collaboratively generated molecular, genetic and digital resources available for it. Here we complement the approximately 140,000 neuron FlyWire whole-brain connectome 1 with a systematic and hierarchical annotation of neuronal classes, cell types and developmental units (hemilineages). Of 8,453 annotated cell types, 3,643 were previously proposed in the partial hemibrain connectome 2 , and 4,581 are new types, mostly from brain regions outside the hemibrain subvolume. Although nearly all hemibrain neurons could be matched morphologically in FlyWire, about one-third of cell types proposed for the hemibrain could not be reliably reidentified. We therefore propose a new definition of cell type as groups of cells that are each quantitatively more similar to cells in a different brain than to any other cell in the same brain, and we validate this definition through joint analysis of FlyWire and hemibrain connectomes. Further analysis defined…

Abstract The increasing demands for more efficient and brighter thin-film light-emitting diodes (LEDs) in flat-panel display and solid-state lighting applications have promoted research into three-dimensional (3D) perovskites. These materials exhibit high charge mobilities and low quantum efficiency droop 1–6 , making them promising candidates for achieving efficient LEDs with enhanced brightness. To improve the efficiency of LEDs, it is crucial to minimize nonradiative recombination while promoting radiative recombination. Various passivation strategies have been used to reduce defect densities in 3D perovskite films, approaching levels close to those of single crystals 3 . However, the slow radiative (bimolecular) recombination has limited the photoluminescence quantum efficiencies (PLQEs) of 3D perovskites to less than 80% (refs. 1,3 ), resulting in external quantum efficiencies (EQEs) of LED devices of less than 25%. Here we present a dual-additive crystallization method that enables the formation of highly efficient 3D perovskites, achieving an exceptional PLQE of 96%. This…

Abstract Floods are one of the most common natural disasters, with a disproportionate impact in developing countries that often lack dense streamflow gauge networks 1 . Accurate and timely warnings are critical for mitigating flood risks 2 , but hydrological simulation models typically must be calibrated to long data records in each watershed. Here we show that artificial intelligence-based forecasting achieves reliability in predicting extreme riverine events in ungauged watersheds at up to a five-day lead time that is similar to or better than the reliability of nowcasts (zero-day lead time) from a current state-of-the-art global modelling system (the Copernicus Emergency Management Service Global Flood Awareness System). In addition, we achieve accuracies over five-year return period events that are similar to or better than current accuracies over one-year return period events. This means that artificial intelligence can provide flood warnings earlier and over larger and more impactful events in ungauged basins. The model developed here was incorporated into an operational early…

Abstract The immune system has a critical role in orchestrating tissue healing. As a result, regenerative strategies that control immune components have proved effective 1,2 . This is particularly relevant when immune dysregulation that results from conditions such as diabetes or advanced age impairs tissue healing following injury 2,3 . Nociceptive sensory neurons have a crucial role as immunoregulators and exert both protective and harmful effects depending on the context 4–12 . However, how neuro–immune interactions affect tissue repair and regeneration following acute injury is unclear. Here we show that ablation of the Na V 1.8 nociceptor impairs skin wound repair and muscle regeneration after acute tissue injury. Nociceptor endings grow into injured skin and muscle tissues and signal to immune cells through the neuropeptide calcitonin gene-related peptide (CGRP) during the healing process. CGRP acts via receptor activity-modifying protein 1 (RAMP1) on neutrophils, monocytes and macrophages to inhibit recruitment, accelerate death, enhance efferocytosis and polarize macrophages…

. Here we show that a peripheral immune insult strongly activates the body-brain axis to regulate immune responses. We demonstrate that pro-inflammatory and anti-inflammatory cytokines communicate with distinct populations of vagal neurons to inform the brain of an emerging inflammatory response. In turn, the brain tightly modulates the course of the peripheral immune response. Genetic silencing of this body-brain circuit produced unregulated and out-of-control inflammatory responses. By contrast, activating, rather than silencing, this circuit affords neural control of immune responses. We used single-cell RNA sequencing, combined with functional imaging, to identify the circuit components of this neuroimmune axis, and showed that its selective manipulation can effectively suppress the pro-inflammatory response while enhancing an anti-inflammatory state. The brain-evoked transformation of the course of an immune response offers new possibilities in the modulation of a wide range of immune disorders, from autoimmune diseases to cytokine storm and shock.

Abstract RAS oncogenes (collectively NRAS , HRAS and especially KRAS ) are among the most frequently mutated genes in cancer, with common driver mutations occurring at codons 12, 13 and 61 1 . Small molecule inhibitors of the KRAS(G12C) oncoprotein have demonstrated clinical efficacy in patients with multiple cancer types and have led to regulatory approvals for the treatment of non-small cell lung cancer 2,3 . Nevertheless, KRAS G12C mutations account for only around 15% of KRAS -mutated cancers 4,5 , and there are no approved KRAS inhibitors for the majority of patients with tumours containing other common KRAS mutations. Here we describe RMC-7977, a reversible, tri-complex RAS inhibitor with broad-spectrum activity for the active state of both mutant and wild-type KRAS, NRAS and HRAS variants (a RAS(ON) multi-selective inhibitor). Preclinically, RMC-7977 demonstrated potent activity against RAS-addicted tumours carrying various RAS genotypes, particularly against cancer models with KRAS codon 12 mutations ( KRAS G12X ). Treatment with RMC-7977 led to tumour regression and was…

In our original article, we provided an extensive description in the Supplementary Information of pseudocode underlying the functioning of our AlphaFold3 model, which allows prediction of the structure of complexes including proteins, nucleic acids, small molecules, ions and modified residues. Accompanying release of the paper we provided free non-commercial access to the AlphaFold3 server ( https://alphafoldserver.com/welcome ). Google DeepMind has now released the underlying inference code, which can be accessed here: https://github.com/google-deepmind/alphafold3 .

Abstract Several theories have been proposed to describe the formation of black hole seeds in the early Universe and to explain the emergence of very massive black holes observed in the first thousand million years after the Big Bang 1–3 . Models consider different seeding and accretion scenarios 4–7 , which require the detection and characterization of black holes in the first few hundred million years after the Big Bang to be validated. Here we present an extensive analysis of the JWST-NIRSpec spectrum of GN-z11, an exceptionally luminous galaxy at z = 10.6, revealing the detection of the [Ne iv ] λ 2423 and CII* λ 1335 transitions (typical of active galactic nuclei), as well as semi-forbidden nebular lines tracing gas densities higher than 10 9 cm −3 , typical of the broad line region of active galactic nuclei. These spectral features indicate that GN-z11 hosts an accreting black hole. The spectrum also reveals a deep and blueshifted CIV λ 1549 absorption trough, tracing an outflow with velocity 800−1,000 km s −1 , probably driven by the active galactic nucleus. Assuming local…

Abstract The glymphatic movement of fluid through the brain removes metabolic waste 1–4 . Noninvasive 40 Hz stimulation promotes 40 Hz neural activity in multiple brain regions and attenuates pathology in mouse models of Alzheimer’s disease 5–8 . Here we show that multisensory gamma stimulation promotes the influx of cerebrospinal fluid and the efflux of interstitial fluid in the cortex of the 5XFAD mouse model of Alzheimer’s disease. Influx of cerebrospinal fluid was associated with increased aquaporin-4 polarization along astrocytic endfeet and dilated meningeal lymphatic vessels. Inhibiting glymphatic clearance abolished the removal of amyloid by multisensory 40 Hz stimulation. Using chemogenetic manipulation and a genetically encoded sensor for neuropeptide signalling, we found that vasoactive intestinal peptide interneurons facilitate glymphatic clearance by regulating arterial pulsatility. Our findings establish novel mechanisms that recruit the glymphatic system to remove brain amyloid.

Abstract The cost of drug discovery and development is driven primarily by failure, with just ∼10% of clinical programs eventually receiving approval. We previously estimated that human genetic evidence doubles the success rate from clinical development to approval. In this study we leverage the growth in genetic evidence over the past decade to better understand the characteristics that distinguish clinical success and failure. We estimate the probability of success for drug mechanisms with genetic support is 2.6 times greater than those without. This relative success varies among therapy areas and development phases, and improves with increasing confidence in the causal gene, but is largely unaffected by genetic effect size, minor allele frequency, or year of discovery. These results suggest we are far from reaching peak genetic insights to aid the discovery of targets for more effective drugs.

Abstract Negotiations for a Global Treaty on plastic pollution1 will shape policy on plastics production, use, and waste management for the coming decades. Parties will require a detailed baseline of waste flows and plastic emission sources at high resolution to enable identification of pollution hotspots and their causes2. Nationally aggregated waste management data can be distributed to smaller scale to identify generalised points of plastic accumulation and source phenomena3-11. However, it is challenging to use this type of spatial allocation to assess the conditions under which emissions take place12,13. To this, we developed a novel methodology for a global macroplastic waste inventory; creating an explanatory framework that combines conceptual modelling of emission mechanisms with measurable activity data. Using machine learning and probabilistic material flow analysis we identify hotspots worldwide (50,072 municipalities) from five land-based plastic waste emission sources. We estimate global plastic waste emissions at 52.5 million metric tonnes (Mt), of which 57% wt. are…

Abstract Emerging spatial computing systems seamlessly superimpose digital information on the physical environment observed by a user, enabling transformative experiences across various domains, such as entertainment, education, communication and training 1–3 . However, the widespread adoption of augmented-reality (AR) displays has been limited due to the bulky projection optics of their light engines and their inability to accurately portray three-dimensional (3D) depth cues for virtual content, among other factors 4,5 . Here we introduce a holographic AR system that overcomes these challenges using a unique combination of inverse-designed full-colour metasurface gratings, a compact dispersion-compensating waveguide geometry and artificial-intelligence-driven holography algorithms. These elements are co-designed to eliminate the need for bulky collimation optics between the spatial light modulator and the waveguide and to present vibrant, full-colour, 3D AR content in a compact device form factor. To deliver unprecedented visual quality with our prototype, we develop an innovative…

. Here we report a single-cell transcriptomic atlas of six different brain regions in the aged human brain, covering 1.3 million cells from 283 post-mortem human brain samples across 48 individuals with and without Alzheimer's disease. We identify 76 cell types, including region-specific subtypes of astrocytes and excitatory neurons and an inhibitory interneuron population unique to the thalamus and distinct from canonical inhibitory subclasses. We identify vulnerable populations of excitatory and inhibitory neurons that are depleted in specific brain regions in Alzheimer's disease, and provide evidence that the Reelin signalling pathway is involved in modulating the vulnerability of these neurons. We develop a scalable method for discovering gene modules, which we use to identify cell-type-specific and region-specific modules that are altered in Alzheimer's disease and to annotate transcriptomic differences associated with diverse pathological variables. We identify an astrocyte program that is associated with cognitive resilience to Alzheimer's disease pathology, tying choline…

Abstract Carbapenem-resistant Acinetobacter baumannii (CRAB) has emerged as a major global pathogen with limited treatment options 1 . No new antibiotic chemical class with activity against A. baumannii has reached patients in over 50 years 1 . Here we report the identification and optimization of tethered macrocyclic peptide (MCP) antibiotics with potent antibacterial activity against CRAB. The mechanism of action of this molecule class involves blocking the transport of bacterial lipopolysaccharide from the inner membrane to its destination on the outer membrane, through inhibition of the LptB 2 FGC complex. A clinical candidate derived from the MCP class, zosurabalpin (RG6006), effectively treats highly drug-resistant contemporary isolates of CRAB both in vitro and in mouse models of infection, overcoming existing antibiotic resistance mechanisms. This chemical class represents a promising treatment paradigm for patients with invasive infections due to CRAB, for whom current treatment options are inadequate, and additionally identifies LptB 2 FGC as a tractable target for…

Abstract Working from home has become standard for employees with a university degree. The most common scheme, which has been adopted by around 100 million employees in Europe and North America, is a hybrid schedule, in which individuals spend a mix of days at home and at work each week 1,2 . However, the effects of hybrid working on employees and firms have been debated, and some executives argue that it damages productivity, innovation and career development 3–5 . Here we ran a six-month randomized control trial investigating the effects of hybrid working from home on 1,612 employees in a Chinese technology company in 2021–2022. We found that hybrid working improved job satisfaction and reduced quit rates by one-third. The reduction in quit rates was significant for non-managers, female employees and those with long commutes. Null equivalence tests showed that hybrid working did not affect performance grades over the next two years of reviews. We found no evidence for a difference in promotions over the next two years overall, or for any major employee subgroup. Finally, null…

Abstract Proving mathematical theorems at the olympiad level represents a notable milestone in human-level automated reasoning 1–4 , owing to their reputed difficulty among the world’s best talents in pre-university mathematics. Current machine-learning approaches, however, are not applicable to most mathematical domains owing to the high cost of translating human proofs into machine-verifiable format. The problem is even worse for geometry because of its unique translation challenges 1,5 , resulting in severe scarcity of training data. We propose AlphaGeometry, a theorem prover for Euclidean plane geometry that sidesteps the need for human demonstrations by synthesizing millions of theorems and proofs across different levels of complexity. AlphaGeometry is a neuro-symbolic system that uses a neural language model, trained from scratch on our large-scale synthetic data, to guide a symbolic deduction engine through infinite branching points in challenging problems. On a test set of 30 latest olympiad-level problems, AlphaGeometry solves 25, outperforming the previous best method that…

Abstract A key challenge in realizing practical quantum networks for long-distance quantum communication involves robust entanglement between quantum memory nodes connected by fibre optical infrastructure1–3. Here we demonstrate a two-node quantum network composed of multi-qubit registers based on silicon-vacancy (SiV) centres in nanophotonic diamond cavities integrated with a telecommunication fibre network. Remote entanglement is generated by the cavity-enhanced interactions between the electron spin qubits of the SiVs and optical photons. Serial, heralded spin-photon entangling gate operations with time-bin qubits are used for robust entanglement of separated nodes. Long-lived nuclear spin qubits are used to provide second-long entanglement storage and integrated error detection. By integrating efficient bidirectional quantum frequency conversion of photonic communication qubits to telecommunication frequencies (1,350 nm), we demonstrate the entanglement of two nuclear spin memories through 40 km spools of low-loss fibre and a 35-km long fibre loop deployed in the Boston area…

Abstract Perovskite solar cells (PSCs) are among the most promising photovoltaic technologies owing to their exceptional optoelectronic properties 1,2 . However, the lower efficiency, poor stability and reproducibility issues of large-area PSCs compared with laboratory-scale PSCs are notable drawbacks that hinder their commercialization 3 . Here we report a synergistic dopant-additive combination strategy using methylammonium chloride (MACl) as the dopant and a Lewis-basic ionic-liquid additive, 1,3-bis(cyanomethyl)imidazolium chloride ([Bcmim]Cl). This strategy effectively inhibits the degradation of the perovskite precursor solution (PPS), suppresses the aggregation of MACl and results in phase-homogeneous and stable perovskite films with high crystallinity and fewer defects. This approach enabled the fabrication of perovskite solar modules (PSMs) that achieved a certified efficiency of 23.30% and ultimately stabilized at 22.97% over a 27.22-cm 2 aperture area, marking the highest certified PSM performance. Furthermore, the PSMs showed long-term operational stability, maintaining…

Abstract Despite tremendous efforts in the past decades, relationships among main avian lineages remain heavily debated without a clear resolution. Discrepancies have been attributed to diversity of species sampled, phylogenetic method and the choice of genomic regions 1–3 . Here we address these issues by analysing the genomes of 363 bird species 4 (218 taxonomic families, 92% of total). Using intergenic regions and coalescent methods, we present a well-supported tree but also a marked degree of discordance. The tree confirms that Neoaves experienced rapid radiation at or near the Cretaceous–Palaeogene boundary. Sufficient loci rather than extensive taxon sampling were more effective in resolving difficult nodes. Remaining recalcitrant nodes involve species that are a challenge to model due to either extreme DNA composition, variable substitution rates, incomplete lineage sorting or complex evolutionary events such as ancient hybridization. Assessment of the effects of different genomic partitions showed high heterogeneity across the genome. We discovered sharp increases in…

profiled the temporal transcriptome, proteome, metabolome, lipidome, phosphoproteome, acetylproteome, ubiquitylproteome, epigenome and immunome in whole blood, plasma and 18 solid tissues in male and female Rattus norvegicus over eight weeks of endurance exercise training. The resulting data compendium encompasses 9,466 assays across 19 tissues, 25 molecular platforms and 4 training time points. Thousands of shared and tissue-specific molecular alterations were identified, with sex differences found in multiple tissues. Temporal multi-omic and multi-tissue analyses revealed expansive biological insights into the adaptive responses to endurance training, including widespread regulation of immune, metabolic, stress response and mitochondrial pathways. Many changes were relevant to human health, including non-alcoholic fatty liver disease, inflammatory bowel disease, cardiovascular health and tissue injury and recovery. The data and analyses presented in this study will serve as valuable resources for understanding and exploring the multi-tissue molecular effects of endurance training…

Highly pathogenic H5N1 avian influenza (HPAI H5N1) viruses occasionally infect, but typically do not transmit, in mammals. In the spring of 2024, an unprecedented outbreak of HPAI H5N1 in bovine herds occurred in the USA, with virus spread within and between herds, infections in poultry and cats, and spillover into humans, collectively indicating an increased public health risk1–4. Here we characterize an HPAI H5N1 virus isolated from infected cow milk in mice and ferrets. Like other HPAI H5N1 viruses, the bovine H5N1 virus spread systemically, including to the mammary glands of both species, however, this tropism was also observed for an older HPAI H5N1 virus isolate. Bovine HPAI H5N1 virus bound to sialic acids expressed in human upper airways and inefficiently transmitted to exposed ferrets (one of four exposed ferrets seroconverted without virus detection). Bovine HPAI H5N1 virus thus possesses features that may facilitate infection and transmission in mammals. HPAI H5N1 virus isolated from infected cow milk is characterized in mice and ferrets, was inefficiently transmitted in…

Decades of neuroimaging studies have shown modest differences in brain structure and connectivity in depression, hindering mechanistic insights or the identification of risk factors for disease onset1. Furthermore, whereas depression is episodic, few longitudinal neuroimaging studies exist, limiting understanding of mechanisms that drive mood-state transitions. The emerging field of precision functional mapping has used densely sampled longitudinal neuroimaging data to show behaviourally meaningful differences in brain network topography and connectivity between and in healthy individuals2–4, but this approach has not been applied in depression. Here, using precision functional mapping and several samples of deeply sampled individuals, we found that the frontostriatal salience network is expanded nearly twofold in the cortex of most individuals with depression. This effect was replicable in several samples and caused primarily by network border shifts, with three distinct modes of encroachment occurring in different individuals. Salience network expansion was stable over time,…

Abstract Reducing body weight to improve metabolic health and related comorbidities is a primary goal in treating obesity 1,2 . However, maintaining weight loss is a considerable challenge, especially as the body seems to retain an obesogenic memory that defends against body weight changes 3,4 . Overcoming this barrier for long-term treatment success is difficult because the molecular mechanisms underpinning this phenomenon remain largely unknown. Here, by using single-nucleus RNA sequencing, we show that both human and mouse adipose tissues retain cellular transcriptional changes after appreciable weight loss. Furthermore, we find persistent obesity-induced alterations in the epigenome of mouse adipocytes that negatively affect their function and response to metabolic stimuli. Mice carrying this obesogenic memory show accelerated rebound weight gain, and the epigenetic memory can explain future transcriptional deregulation in adipocytes in response to further high-fat diet feeding. In summary, our findings indicate the existence of an obesogenic memory, largely on the basis of…

. It remains unclear how human brain network changes relate to subjective and lasting effects of psychedelics. Here we tracked individual-specific brain changes with longitudinal precision functional mapping (roughly 18 magnetic resonance imaging visits per participant). Healthy adults were tracked before, during and for 3 weeks after high-dose psilocybin (25 mg) and methylphenidate (40 mg), and brought back for an additional psilocybin dose 6-12 months later. Psilocybin massively disrupted functional connectivity (FC) in cortex and subcortex, acutely causing more than threefold greater change than methylphenidate. These FC changes were driven by brain desynchronization across spatial scales (areal, global), which dissolved network distinctions by reducing correlations within and anticorrelations between networks. Psilocybin-driven FC changes were strongest in the default mode network, which is connected to the anterior hippocampus and is thought to create our sense of space, time and self. Individual differences in FC changes were strongly linked to the subjective psychedelic…

. We combine satellite imagery, vessel GPS data and deep-learning models to map industrial vessel activities and offshore energy infrastructure across the world's coastal waters from 2017 to 2021. We find that 72-76% of the world's industrial fishing vessels are not publicly tracked, with much of that fishing taking place around South Asia, Southeast Asia and Africa. We also find that 21-30% of transport and energy vessel activity is missing from public tracking systems. Globally, fishing decreased by 12 ± 1% at the onset of the COVID-19 pandemic in 2020 and had not recovered to pre-pandemic levels by 2021. By contrast, transport and energy vessel activities were relatively unaffected during the same period. Offshore wind is growing rapidly, with most wind turbines confined to small areas of the ocean but surpassing the number of oil structures in 2021. Our map of ocean industrialization reveals changes in some of the most extensive and economically important human activities at sea.

Abstract Genome-wide association analyses using high-throughput metabolomics platforms have led to novel insights into the biology of human metabolism1–7. This detailed knowledge of the genetic determinants of systemic metabolism has been pivotal for uncovering how genetic pathways influence biological mechanisms and complex diseases8–11. Here we present a genome-wide association study for 233 circulating metabolic traits quantified by nuclear magnetic resonance spectroscopy in up to 136,016 participants from 33 cohorts. We identify more than 400 independent loci and assign probable causal genes at two-thirds of these using manual curation of plausible biological candidates. We highlight the importance of sample and participant characteristics that can have significant effects on genetic associations. We use detailed metabolic profiling of lipoprotein- and lipid-associated variants to better characterize how known lipid loci and novel loci affect lipoprotein metabolism at a granular level. We demonstrate the translational utility of comprehensively phenotyped molecular data,…

Abstract The first observations of the James Webb Space Telescope (JWST) have revolutionized our understanding of the Universe by identifying galaxies at redshift z ≈ 13 (refs. 1–3 ). In addition, the discovery of many luminous galaxies at Cosmic Dawn ( z > 10) has suggested that galaxies developed rapidly, in apparent tension with many standard models 4–8 . However, most of these galaxies lack spectroscopic confirmation, so their distances and properties are uncertain. Here we present JWST Advanced Deep Extragalactic Survey–Near-Infrared Spectrograph spectroscopic confirmation of two luminous galaxies at $$z={14.32}_{-0.20}^{+0.08}$$ z = 14.32 − 0.20 + 0.08 and z = 13.90 ± 0.17. The spectra reveal ultraviolet continua with prominent Lyman-α breaks but no detected emission lines. This discovery proves that luminous galaxies were already in place 300 million years after the Big Bang and are more common than what was expected before JWST. The most distant of the two galaxies is unexpectedly luminous and is spatially resolved with a radius of 260 parsecs. Considering also the very…

Abstract Type 2 diabetes mellitus is a major risk factor for hepatocellular carcinoma (HCC). Changes in extracellular matrix (ECM) mechanics contribute to cancer development 1,2 , and increased stiffness is known to promote HCC progression in cirrhotic conditions 3,4 . Type 2 diabetes mellitus is characterized by an accumulation of advanced glycation end-products (AGEs) in the ECM; however, how this affects HCC in non-cirrhotic conditions is unclear. Here we find that, in patients and animal models, AGEs promote changes in collagen architecture and enhance ECM viscoelasticity, with greater viscous dissipation and faster stress relaxation, but not changes in stiffness. High AGEs and viscoelasticity combined with oncogenic β-catenin signalling promote HCC induction, whereas inhibiting AGE production, reconstituting the AGE clearance receptor AGER1 or breaking AGE-mediated collagen cross-links reduces viscoelasticity and HCC growth. Matrix analysis and computational modelling demonstrate that lower interconnectivity of AGE-bundled collagen matrix, marked by shorter fibre length and…

Abstract Ion-conducting polymer membranes are essential in many separation processes and electrochemical devices, including electrodialysis 1 , redox flow batteries 2 , fuel cells 3 and electrolysers 4,5 . Controlling ion transport and selectivity in these membranes largely hinges on the manipulation of pore size. Although membrane pore structures can be designed in the dry state 6 , they are redefined upon hydration owing to swelling in electrolyte solutions. Strategies to control pore hydration and a deeper understanding of pore structure evolution are vital for accurate pore size tuning. Here we report polymer membranes containing pendant groups of varying hydrophobicity, strategically positioned near charged groups to regulate their hydration capacity and pore swelling. Modulation of the hydrated micropore size (less than two nanometres) enables direct control over water and ion transport across broad length scales, as quantified by spectroscopic and computational methods. Ion selectivity improves in hydration-restrained pores created by more hydrophobic pendant groups. These…

Abstract Prime editing enables the precise modification of genomes through reverse transcription of template sequences appended to the 3′ ends of CRISPR–Cas guide RNAs 1 . To identify cellular determinants of prime editing, we developed scalable prime editing reporters and performed genome-scale CRISPR-interference screens. From these screens, a single factor emerged as the strongest mediator of prime editing: the small RNA-binding exonuclease protection factor La. Further investigation revealed that La promotes prime editing across approaches (PE2, PE3, PE4 and PE5), edit types (substitutions, insertions and deletions), endogenous loci and cell types but has no consistent effect on genome-editing approaches that rely on standard, unextended guide RNAs. Previous work has shown that La binds polyuridine tracts at the 3′ ends of RNA polymerase III transcripts 2 . We found that La functionally interacts with the 3′ ends of polyuridylated prime editing guide RNAs (pegRNAs). Guided by these results, we developed a prime editor protein (PE7) fused to the RNA-binding, N-terminal domain of…

Abstract Interleukin-10 (IL-10) is a key anti-inflammatory cytokine that can limit immune cell activation and cytokine production in innate immune cell types 1 . Loss of IL-10 signalling results in life-threatening inflammatory bowel disease in humans and mice—however, the exact mechanism by which IL-10 signalling subdues inflammation remains unclear 2–5 . Here we find that increased saturated very long chain (VLC) ceramides are critical for the heightened inflammatory gene expression that is a hallmark of IL-10 deficiency. Accordingly, genetic deletion of ceramide synthase 2 (encoded by Cers2 ), the enzyme responsible for VLC ceramide production, limited the exacerbated inflammatory gene expression programme associated with IL-10 deficiency both in vitro and in vivo. The accumulation of saturated VLC ceramides was regulated by a decrease in metabolic flux through the de novo mono-unsaturated fatty acid synthesis pathway. Restoring mono-unsaturated fatty acid availability to cells deficient in IL-10 signalling limited saturated VLC ceramide production and the associated…

Abstract Chimeric antigen receptor (CAR) T cell therapy has transformed the treatment of haematological malignancies such as acute lymphoblastic leukaemia, B cell lymphoma and multiple myeloma 1–4 , but the efficacy of CAR T cell therapy in solid tumours has been limited 5 . This is owing to a number of factors, including the immunosuppressive tumour microenvironment that gives rise to poorly persisting and metabolically dysfunctional T cells. Analysis of anti-CD19 CAR T cells used clinically has shown that positive treatment outcomes are associated with a more ‘stem-like’ phenotype and increased mitochondrial mass 6–8 . We therefore sought to identify transcription factors that could enhance CAR T cell fitness and efficacy against solid tumours. Here we show that overexpression of FOXO1 promotes a stem-like phenotype in CAR T cells derived from either healthy human donors or patients, which correlates with improved mitochondrial fitness, persistence and therapeutic efficacy in vivo. This work thus reveals an engineering approach to genetically enforce a favourable metabolic…

Abstract The need for improved functionalities in extreme environments is fuelling interest in high-entropy ceramics 1–3 . Except for the computational discovery of high-entropy carbides, performed with the entropy-forming-ability descriptor 4 , most innovation has been slowly driven by experimental means 1–3 . Hence, advancement in the field needs more theoretical contributions. Here we introduce disordered enthalpy–entropy descriptor (DEED), a descriptor that captures the balance between entropy gains and enthalpy costs, allowing the correct classification of functional synthesizability of multicomponent ceramics, regardless of chemistry and structure. To make our calculations possible, we have developed a convolutional algorithm that drastically reduces computational resources. Moreover, DEED guides the experimental discovery of new single-phase high-entropy carbonitrides and borides. This work, integrated into the AFLOW computational ecosystem, provides an array of potential new candidates, ripe for experimental discoveries.