Browse Articles

An embedding-based framework enables statistical testing of gene-set function hypotheses inferred by large language models

Abstract Emerging large language models (LLMs) can infer gene functions directly from gene lists, enabling hypothesis generation without predefined gene sets. However, these LLM-derived predictions are qualitative, and principled statistical validation is lacking. Here, we develop an embedding-based statistical framework that transforms gene and function descriptions into vector representations, enabling statistical testing of gene-gene and gene-function relationships and quantitative prioritization of de novo functional hypotheses inferred by LLMs. We benchmark seven state-of-the-art embedding models using curated and retrieval-augmented literature-derived gene descriptions across diverse biological contexts. OpenAI’s text-embedding-3-large and Google’s gemini-embedding-001 perform best, capturing gene-gene functional relationships in 88.7-92.5% of Gene Ontology biological processes and approximately 98.6% of canonical pathways. In gene-function association analyses, these models achieve high sensitivity (95.2-98.4%) and specificity (72.7-84.3%). Through contamination analysis and evaluation using experimentally informed protein assembly gene sets, our framework distinguishes biologically meaningful LLM-inferred hypotheses from noise, outperforming confidence-based inference and conventional enrichment analysis. We further develop the open-source R package DEGEmbedR and demonstrate its utility for interpreting a drug perturbation-derived differentially expressed gene (DEG) signature lacking significant conventional enrichment results. Together, these results establish LLM-derived embeddings as a quantitative foundation for functional genomics and the statistical validation of LLM-based gene function inference.

Synergistic Steric Hindrance and Chlorination Enable Efficient Binary Organic Solar Cells with Low Energy Loss

A Wireless, Passive Multimodal Wearable Sensor With Decoupled Sensing Capabilities

ABSTRACT Developing electronic skin (e‐skin) that perceives multi‐stimuli and even exceeds human skin sensing capacity remains a major challenge. Yet existing technologies focused mainly on tactile sensing suffer from poor superimposed multi‐stimuli discrimination, external power dependence, and wired transmission modes. Here, we report an integrated wireless passive wearable sensor based on frequency‐division inductance‐capacitance (LC) resonator array, capable of simultaneously detecting superimposed multi‐stimuli including pressure, odor and humidity, along with decoupling. The system's performance is designed and validated using three‐dimensional full‐wave electromagnetic simulations. Notably, pressure‐sensing with an ultrafast response/recovery (∼5/6 ms) and high sensitivity (6.15 MHz·kPa −1 ) is enabled by a gradient‐modulus trilayer hydrogel incorporating a micro‐pyramidal patterned top‐layer. Furthermore, the sensor demonstrates trace‐level (200 ppb) NO 2 detection without interference from humidity or pressure and exhibits high humidity sensitivity across a wide humidity range (2%–98% RH). Demonstration of this sensor as e‐skin reveals capabilities surpassing previous devices, enabling wireless passive and decoupled detection of small applied mechanical pressure, trace‐level NO 2 , and ambient humidity under complex stimuli conditions, showing high selectivity and minimal cross‐sensitivity. The proposed system introduces a transformative approach, unlocking substantial benefits for a variety of wearable applications.

The basolateral amygdala decodes visceral pain and itch via distinct circuits and molecules

Origin of Dark Current Robustness in Photomultiplication Organic Photodetectors Enabled by Ionic Conjugated Electron‐Blocking Layers

ABSTRACT Photomultiplication‐type organic photodetectors (PM OPDs), which rely on signal amplification via electron trapping, suffer from a fundamental trade‐off: high gain is typically accompanied by a steep increase in dark current under high reverse bias. Here, we overcome this challenge by employing ionic conjugated polyelectrolyte (CPE)‐based electron blocking layers (EBLs) with an alternating fluorene‐co‐triphenylamine backbone. CPE‐based EBLs in PM OPDs enable robust dark current stability under reverse bias up to −10 V while maintaining external quantum efficiencies (EQE) exceeding 2000%. Charge‐dynamics analysis using Fowler–Nordheim plots reveals that a nanometer‐thick CPE layer provides electron‐blocking performance comparable to that of a ∼20 nm Al 2 O 3 layer. At the same time, we show that the nature of the ionic side chains (cationic vs anionic) in CPEs governs interfacial energy‐level alignment, thereby modulating hole selectivity and photocarrier dynamics. The design principles establish a general framework for interfacial and field engineering across a broad range of photodetector platforms—from organic to hybrid systems—guiding the development of next‐generation photodetectors.

The m⁶A reader YTHDF2 regulates mRNA transport and axon outgrowth in developing neurons

Controlled Synthesis of High‐Nuclearity Th ₃₈ Cluster for Efficient Photocatalytic Dehydrogenative Cross‐Coupling

ABSTRACT Actinide clusters represent a significant platform for decoding the complex chemical behavior of actinide elements, providing significant insights for the development of nuclear reactors and the management of radioactive waste. However, compared to other metal clusters, actinide clusters are still underexplored, especially high‐nuclearity thorium clusters. Here, we developed a synthetic strategy, in the form of “cluster from cluster” to construct the highest nuclearity thorium cluster ( Th 38 ) so far. It exhibits remarkable stability under high γ‐ray irradiation (up to 690 kGy) and displays unprecedented dispersibility and stability under strong acid conditions (pH = 1). Moreover, the high stability and unique electronic structure of Th 38 renders it an excellent catalyst for the photocatalytic synthesis of 4‐quinazolinones. This work utilizes a strategy to achieve the selfassembly of the highest nuclearity Th 38 and expands its potential applications.

Uranium in cobalt-hydroxide exports from the Democratic Republic of the Congo

Abstract The Democratic Republic of Congo (DRC)’s Copperbelt is the world’s leading source of cobalt, a metal essential to the energy transition. In this region, uranium co-occurs with oxidized cobalt ores. Despite no officially reported uranium production, we show that uranium is extracted, concentrated, and exported via the DRC cobalt supply chain. This conclusion follows from co-mineralization, similar behavior of cobalt and uranium in hydrometallurgical circuits, and the partitioning of uranium into crude cobalt hydroxide, the export form for over 95% of DRC production. We combine a first-order carryover model with geological mapping, mineralization and geochemical data, as well as mine-level trade records (2000–2024) to quantify these flows. We estimate 2,000–5,000 tonnes of natural uranium were exported embedded in cobalt-hydroxide shipments. Less than 10% of this material has been publicly declared and placed under international safeguards. In addition, 1,000–4,000 tonnes were likely discarded to tailings in easily-mobilized forms, posing environmental and health risks. Together, these findings identify a substantial gap in nuclear accountancy and environmental oversight within a critical mineral supply chain.

Ultrathin Hydrogel Membranes Inspired by Soap Films Enable Physiologically Relevant Breathing Lung Models

ABSTRACT Breathing continuously stretches the lungs, providing essential mechanical cues that regulate cellular behavior and disease responses. Mimicking the repetitive out‐of‐plane deformation of alveolar tissues driven by transpulmonary pressure requires ultrathin stretchable hydrogel membranes that sustain prolonged cyclic loading in hydrated environments. However, membrane thinning inevitably amplifies stress concentration and fatigue failure, making the simultaneous achievement of ultrathin geometry and long‐term durability fundamentally incompatible in conventional hydrogel systems. Here, we present a soap film‐inspired hydrogel membrane that overcomes this trade‐off and enables dynamic breathing lung models. By incorporating acrylamide into gelatin methacryloyl, precursor viscosity and surface tension are systematically tuned to generate ultrathin liquid films governed by Frankel's law, while a hybrid co‐polymer network is formed to produce durable, freestanding membranes. Integrated with a negative‐pressure bioreactor that reproduces physiological breathing under air‐liquid interface culture (10%–15% strain, 12 cycles min −1 ), the platform establishes a dynamic lung model with physiologically relevant out‐of‐plane deformation. Cyclic breathing activates YAP‐mediated mechanotransduction, drives tissue and extracellular matrix remodeling, and modulates inflammatory and antiviral responses in an influenza A infection model. This work establishes a physics‐guided strategy for engineering ultrathin durable hydrogel membranes and advances mechanically faithful in vitro lung models for mechanobiology and respiratory disease research.

Effect of a complementary feeding intervention based on iron- and zinc-biofortified pearl millet on the gut microbiota in 12–18-month-old children: a randomized trial

Asymmetric Lamellar Templating of the Perovskite/C ₆₀ Interface for Scalable Inverted Perovskite Photovoltaics

ABSTRACT Scalable inverted perovskite solar cells require top interfaces that combine defect passivation with spatially uniform electron extraction as the device area increases. Low‐dimensional capping layers based on bulky spacer cations can suppress surface recombination, but poorly controlled molecular packing often introduces transport barriers and local interfacial heterogeneity. Here, we introduce indol‐3‐ylethylammonium iodide (Ind) as a π‐electron‐rich spacer cation for asymmetric lamellar templating at the perovskite/C 60 junction. The heteroatom‐polarized indole framework creates lateral electrostatic anisotropy within the aromatic plane, promoting face‐to‐face spacer association and aligning the lamellar interphase. The resulting π‐rich lamellar interphase suppresses interfacial recombination, preserves electron extraction, and improves spatial optoelectronic uniformity in large‐area devices. Ind‐based devices achieve a certified efficiency of 26.94% in small‐area cells and a certified module efficiency of 23.21% for a 25 cm 2 monolithic module. They further retain > 93% efficiency after 1000 h at 85°C and maintain 85% of their initial efficiency after 1800 h under continuous 1‐sun MPP tracking. These results identify lateral electrostatic anisotropy as a molecular design principle for scalable perovskite/C 60 top interfaces.

Gene regulatory innovations from transposable elements in primate cerebellum development

Abstract Transposable elements are hypothesized to have driven gene regulatory innovation, yet their contributions to primate brain development at the cell type level remain underexplored. Here, we use single-cell multiomics data from human, macaque, marmoset, and mouse cerebella to show that transposable element contributions to different cell types are shaped by varying degrees of constraints across cell types, as well as the preferential co-option of certain transposable elements in specific cell states. Using a sequence-based deep-learning model that predicts cell-type-specific chromatin accessibility, we systematically assess the co-option potential of transposable elements into cerebellar gene regulatory networks, identifying twelve transposable element subfamilies with complex regulatory sequences in their ancestral states that facilitate their co-option as cell-type-specific cis-regulatory elements. Preservation of these ancestral regulatory sequences, as well as the active chromatin environment surrounding the insertion site, is the major determinant of the accessibility of extant copies. Lineage-specific accessible copies contribute to human-specific gene expression. Broadly, we demonstrate how transposable elements can be flexibly co-opted into cell-type-specific gene regulatory networks, and introduce a generalizable analytical framework for dissecting their contribution to mammalian regulatory evolution.

Reconfigurable Ferroelectric‐Like Two‐Dimensional Electron Gas at Room Temperature

ABSTRACT Engineering ferroelectric two‐dimensional electron gases (2DEGs) offers a promising route toward nonvolatile and reconfigurable electronic functionalities, yet room temperature implementations remain scarce. Here, we report a room temperature ferroelectric‐like 2DEG at the SrO‐ interface between (001) (STO) and (KNN) thin films. Robust ferroelectricity is confirmed by piezoforce microscopy, optical second harmonic generation, Nb‐/Ti‐ edges x‐ray linear dichroism, scanning transmission electron microscopy, and macroscopic polarization measurements. The ferroelectric polarization enables nonvolatile control of interfacial transport and the realization of reconfigurable conducting channels with arbitrary geometry at room temperature. The coupling between ferroelectric order and 2DEG conductivity arises from the extension of the polarization into the interfacial unit cells hosting the electron gas. These results establish / heterostructures as a ferroelectric‐like oxide electronic platform, opening pathways for low‐power, nonvolatile electronic and spin‐orbitronic devices.

Tunable photofluorochromic switch driven by dynamic through-space conjugation

Mechanically Programmable Electromagnetic Metamaterials for Generalized Phase Tailoring With Zero Static Power Consumption

ABSTRACT Mechanically modulated reconfigurable electromagnetic metamaterials represent a promising avenue for flexible wavefront manipulation. However, most mechanically tunable designs rely on collective deformations and continuous external loading, leading to limited programmability and high static power consumption. Here, we present a mechanically programmable electromagnetic metamaterial enabled by 3D‐printed shape memory polymer (SMP) compression‐torsion coupling structures integrated with the three‐fold symmetric three‐armed meta‐atoms (C3 meta‐atoms) for generalized phase tailoring with zero static power consumption. The compression‐torsion coupling structures enable deterministic and independent in‐plane rotation of each unit cell under vertical compression, while the C3 meta‐atoms provide sixfold cross‐circularly polarized phase amplification, achieving full 0°–360° phase coverage with a narrow rotational angular range of 0°–60°. Leveraging the intrinsic shape‐locking and shape‐recovery properties of SMP, arbitrary phase distribution patterns are attainable via mechanical coding without sustained power consumption, and can be repeatedly erased and rewritten via thermal recovery. Numerical simulations and experimental characterizations reveal the design principle and operation mechanism of the metamaterial, verifying its programmable functionalities through demonstrations of anomalous refraction, reconfigurable metalens, and orbital‐angular‐momentum (OAM) generators. These findings provide a conceptual framework for low‐energy, programmable, and reconfigurable wavefront modulation, laying a foundation for advancing next‐generation mechanically programmable electromagnetic metamaterials.

Spatiotemporal patterns of lower-band whistler mode waves in the magnetosphere of Earth

Abstract Electromagnetic emissions of lower-band chorus and exohiss waves are known to interact with electrons in the Earth’s outer radiation belt. Understanding statistical properties of these natural emissions is essential for modeling space weather hazards and protecting satellite infrastructure. Despite extensive previous studies, their spatiotemporal characteristics remain incompletely understood. Here, we analyze a large data set of spacecraft measurements, showing that distributions of wave amplitudes exhibit a wide log-normal core with an irregular heavy tail, changing with geomagnetic activity and location. The distribution of time intervals between wave detections follows a power law, suggestive of temporal bunching. Around local noon, we nearly always detect waves up to high latitudes, while intense waves occur for less than a few percent of the time, predominantly in the postmidnight equatorial region. Our findings indicate that the burstiness of whistler-mode waves may not be fully captured by long-term averages, which are commonly used in radiation belt modeling.

Corneal allogenic intrastromal ring segments for the treatment of keratoconus

Quaternary PdPS _0.55 Se _0.45 for SBUV‐to‐SWIR Broadband Photodetection and Tri‐Band In‐Sensor Processing

ABSTRACT Post‐fire remote sensing for accurate damage assessment critically relies on multi‐band perception spanning from solar‐blind ultraviolet (SBUV) to short‐wave infrared (SWIR) and effective in‐sensor image pre‐processing. However, most vdW photodetectors still operate over limited spectral response, which constrains unified ultraviolet‐visible‐infrared sensing and computing within a single device platform. Here, we develop a stable quaternary vdW semiconductor, PdPS 0.55 Se 0.45 , and demonstrate a single photodetector enabling broadband sensing from SBUV (266 nm) to SWIR (1550 nm) while supporting in‐sensor convolutional processing for remote‐sensing images. The device achieves a peak responsivity ( R ) of 98.13 (84.81) A W −1 and a specific detectivity ( D *) exceeding 10 13 Jones at 266 nm (638 nm). We further exploit the intrinsic power‐density‐dependent responsivity to program band‐specific convolution kernels, where responsivity differences under 266, 638, and 1550 nm illumination are mapped into analog multiply‐accumulate weights. Coupled with a convolutional neural network (CNN), this tri‐band in‐sensor pre‐processing enables robust post‐fire target recognition on noise‐corrupted remote sensing images, achieving a recognition accuracy of ∼96% for post‐fire scenes. This work offers a practical route to SBUV‐to‐SWIR photodetector for in‐sensor computing, advancing broadband perception‐computation integration and creating new opportunities for remote sensing vision under complex environments.

ERH elicits cell lineage restriction in mammalian preimplantation development and differentiation from pluripotency via H3K9me3-heterochromatin

Extracellular vesicle-derived candidate biomarker in metabolic dysfunction-associated steatotic liver disease: systematic review and exploratory meta-analysis

Abstract Metabolic dysfunction-Associated Steatotic Liver Disease (MASLD) is raising global concern, highlighting the need for simple, reliable, and non-invasive diagnostic tools. Extracellular vesicles (EVs) are well-known nano-carriers of circulating biomarkers, which are often involved in disease development and intercellular communication. The current review aims to synthesise and analyse the evidence from the exploratory diagnostic studies examining the EV-derived biomarkers in MASLD. A detailed literature search across three electronic databases, including PubMed, Scopus, and Embase, resulted in 3440 records, of which 7 eligible studies were selected for further data synthesis. Quality assessment for the included studies was carried out adhering to Minimal Information for Studies of Extracellular Vesicles (MISEV) guidelines. The meta-analysis was performed using unique analytic strategies, including area under the curve (AUC)-based analyses for EV-derived panels and biomarkers in MASLD. The meta-analysis results indicated that the diagnostic accuracy associated with the plasma-derived EV microRNAs (miRNAs) was considerable for the exploratory studies and needs further validation for clinical practice. Overall, systematically synthesised evidence suggests that the EV-derived miRNAs might serve as clinically useful diagnostic biomarkers for early-stage MASLD. However, the exploratory nature of the literature, which involved high heterogeneity across the study design, sample processing, and analysis, caused difficulty in generalising the conclusions, thus suggesting the need for standardisation. Overall, the present review provides a detailed synthesis of exploratory research, emphasising the diagnostic value of EV-derived biomarkers.