Search

Shortcut learning, where models exploit spurious correlations in data, leads to brittle predictions and undermines fair assessment of model performance. Here, the authors introduce a diagnostic paradigm—shortcut hull learning—that identifies shortcuts from datasets, enabling unbiased evaluation of model capabilities.

The red fluorescent protein mScarlet3-H is bright, photostable and very robust to high temperature, chaotropic conditions and oxidative environments. mScarlet3-H works well in correlative light and electron microscopy, tissue clearing and time-lapse super-resolution microscopy.

Structural and biochemical studies of the monkeypox virus core protease show that it exists as an active homodimer and indicate that the substrate-binding sites of core proteases are a promising target for antiviral drugs.

Photostrictive materials offer a promising solution for the realization of wireless microelectromechanical devices. Here the authors demonstrate a method for developing high-performance inorganic photostrictive materials by constructing a polymorphic phase boundary.

Driving comfort in autonomous vehicles is crucial but often overlooked in global path planning. Here, authors developed a model using multi-head attention and XGBoost to predict driving comfort from road information, reducing jerk by 15% and improving comfort scores by 13% in real vehicle tests.

A method called UltraSelex for identifying RNA aptamers has been developed. UltraSelex is noniterative and can combine biochemical partitioning, high-throughput sequencing and computational signal-to-background rank modeling of hits in approximately 1 day.

Comprehensive analyses of Cas9 proteins shed light on the evolution of the CRISPR–Cas9 system, and identify a pro-CRISPR accessory protein in bacteria that boosts CRISPR-mediated immunity by enhancing the DNA binding and cleavage activity of Cas9.

Doping is a common strategy to improve thermoelectric performance yet limited to the solid solubility of dopants. Here, the authors find that forming metavalently bonded precipitates is key to property enhancement in Te rather than modifying the matrix lattice.

SGLT plays key roles in sugar uptake and reabsorption. Here the authors provided the cryo-EM structures of human SGLT1 and SGLT2 in the substrate-bound occluded state, uncovering the conformational changes of SGLT during sugar transport.

Surface magnetic order could precede bulk order, but detecting such a separation necessitates experimental probes sensitive to both surface and bulk phase transitions. Here, using second harmonic generation, the authors propose that this situation is realized in a van der Waals antiferromagnet CrSBr.

SGLT1 is essential for glucose, galactose and water uptake from the intestine, and its inhibitors have broad therapeutic potential. Here, the authors describe the cryo-EM structure of human SGLT1 in complex with an inhibitor.

Ambiguity in human-oriented traffic laws poses a significant challenge to the regulation of self-driving vehicles. Here, the authors present a trigger-based hierarchical online compliance monitor for self-assessment of self-driving vehicles using ambiguous compliance threshold selection principles.

Climate mitigation have broad effects on achieving Sustainable Development Goals. This paper explores the spillover effects of China’s climate mitigation pathways and identify a cost-effective pathway with co-benefits for sustainability.

The cryo-electron microscopy structure of mouse PIEZO2 is determined and compared to that of PIEZO1, providing insights into the potential gating mechanisms of these mechanosensitive ion channels.

mBaoJin is a monomeric derivative of the bright and photostable green fluorescent protein StayGold. mBaoJin offers favorable photophysical properties for use in diverse protein tagging and subcellular labeling applications.

The molecular differences between the two major gastric cancer subtypes diffuse-type gastric cancer (DGC) and intestinaltype gastric cancer (IGC) remain to be investigated. Here, integrated analysis of proteome, phospho-proteome and transcription factor activity for DGC and IGC reveals potential subtypes.

Clear cell renal cell carcinoma is an aggressive form of renal cancer, with differences in genomic mutations reported between Western and Eastern populations. In this study, the authors have compiled proteogenomic analysis of Chinese ccRCC to reveal genomic alterations and dysregulation of immune and metabolic responses.

High-throughput computation is especially useful for materials screening where synthesis is challenging. Here, it is used to construct a stability map of ternary nitrides, allowing discovery of stable compounds and providing insight into principles that govern nitride stability.

Here the authors combine functional and structural work to show that METTL4 functions as U2 snRNA m6Am RNA methyltransferase supporting N6-methylation of 2’-O-methyladenosine in Arabidopsis.

Carnosine is a mobile buffering metabolite. Here the authors link carnosine accumulation, hypoxia and intracellular pH homeostasis in cancer cells as a mechanism of tumor immune evasion via NFX1 degradation and galectin-9 activity.

Multi-animal behaviour quantification is pivotal for deciphering animal social behaviours and has broad applications in neuroscience and ecology. Han and colleagues develop a few-shot learning framework for multi-animal 3D pose estimation, identity recognition and social behaviour classification.

Exercise during the early active phase leads to more robust bone growth and maturation in young mice, through a mechanism dependent on oxidative phosphorylation.

Here, by characterizing 477 genomes from cultivated bacteria and metagenome-assembled genomes of the bee gut microbiota, the authors uncover uncharacterized biosynthetic gene clusters encoding small molecules with potential antimicrobial activity against bee pathogens, shedding light on the role of microbiome in honeybee health.

It is currently challenging to identify protein structures at low concentrations. Here the authors report optical tweezers-coupled Raman spectroscopy to generate tunable and reproducible SERS enhancements with single-molecule level sensitivity and use the method to detect protein structural features.

Many clear cell renal cell carcinoma (ccRCC) patients do not respond or develop resistance to tyrosine kinase inhibitors, such as Sunitinib. Here, the authors perform a proteogenomics analysis of Chinese ccRCC patients treated with Sunitinib and develop a multi-omics classifier to distinguish responders from non-responders.

GY3 encodes a LOG-like protein that acts as a 5′-ribonucleotide phosphohydrolase in cytokinin synthesis. A Gypsy solo-LTR insertion downregulates GY3 expression and shows the potential to enhance grain yield in rice.

In contrast to the extensively studied mechanosensory hair bundle, the cuticular plate is not as well understood. In this study, authors describe the discovery of a hair cell protein called LIM only protein 7, which is localized in the cuticular plate and the cell junction and may play a role in age-related deafness.

Digital twins can be used to support planetary operations and analysis. Here, the authors show tri-aspect characterization of lunar far side regolith and investigate the origin of a fin-shaped rock via digital twin of Yutu-2 rover.

Chips with 256 × 256 memristor arrays that were monolithically integrated on complementary metal–oxide–semiconductor (CMOS) circuits in a commercial foundry achieved 2,048 conductance levels in individual memristors.

With the example of the paraspeckle protein PSPC1, Shao et al. demonstrated the synergistic interplay of promoter-associated RNA and its binding proteins in promoting transcription condensate formation and Pol II engagement and activity via phase separation.

TCF1⁺ ‘stem-like’ CD4⁺ T cells have a capacity for self-renewal and effector differentiation when required. Here the authors show how these stem-like T cells mediate allograft rejection via the replenishment of their effector differentiation.

Myeloid-derived suppressor cells play an important role in immune suppression caused by Echinococcus multilocularis infection. This study demonstrates that the removal of this cell type effectively restores T-cell function and potentiates PD-1 blockade therapy in chronic liver infection.

CO₂ hydrogenation is promising for the conversion of waste CO₂ emissions into value-added chemicals. This Review examines the atomic-level design of heterogeneous catalysts with precise active sites and related catalytic reaction engineering for tuning CO₂ hydrogenation selectivity.

Cells lacking TET proteins and DNMT3A and DNMT3B show that DNMT1 has imprecise maintenance activity and weak de novo activity leading to spontaneous ‘epimutations’. These epimutations can be corrected through a neighbor-guided mechanism.

The GlycoSCORES method, which involves cell-free protein expression and substrate-site profiling of glycosyltransferase enzymes by SAMDI–MS, enables the identification of glycosylation tags for glycoengineering efforts.

Designing vaccine immunogens is often a tedious process. Here the authors develop a deep mutational scanning-based method to rapidly and comprehensively identify prefusion stabilizing mutations of SARS-CoV-2 spike as a vaccine immunogen.

Structural and biophysical analysis of the histone acetyltransferase MOZ double PHD finger (DPF) ___domain reveal that DPF exhibits strong binding preference for crotonylated Lys14 in histone H3 (H3K14) and are co-localized in cells.

Periplasmic proteases DegP and DegQ along with the disulfide bond isomerase DsbC are implicated in CadC-mediated antiacid response in E. coli.

The complete skeleton of a new avialan dinosaur from the Tiaojishan Formation (Middle–Late Jurassic period) of Liaoning Province, China, is described and included in a comprehensive phylogenetic analysis of basal Paraves.

The activation of TGF-β signaling has been implicated in cancer metastasis. Here, the authors show that OTUD1 suppresses metastasis by antagonizing the TGF-β pathway via the deubiquitination of SMAD7, and its loss correlates with poor prognosis in breast cancer.

Deep neural networks are increasingly popular in data-intensive applications, but are power-hungry. New types of computer chips that are suited to the task of deep learning, such as memristor arrays where data handling and computing take place within the same unit, are required. A well-used deep learning model called long short-term memory, which can handle temporal sequential data analysis, is now implemented in a memristor crossbar array, promising an energy-efficient and low-footprint deep learning platform.

Memristor crossbars with array sizes of up to 128 × 64 cells are capable of analogue vector-matrix multiplication and can be used for signal processing, image compression and convolutional filtering.

Single-cell RNA sequencing is used to generate a dataset covering all major human organs in both adult and fetal stages, enabling comparison with similar datasets for mouse tissues.

Memristor-based neural networks hold promise for neuromorphic computing, yet large-scale experimental execution remains difficult. Here, Xia et al. create a multi-layer memristor neural network with in-situ machine learning and achieve competitive image classification accuracy on a standard dataset.

Memristive devices can provide energy-efficient neural network implementations, but they must be tailored to suit different network architectures. Wang et al. develop a trainable weight-sharing mechanism for memristor-based CNNs and ConvLSTMs, achieving a 75% reduction in weights without compromising accuracy.

Leaky integrate-and-fire artificial neurons based on diffusive memristors enable unsupervised weight updates of drift-memristor synapses in an integrated convolutional neural network capable of pattern recognition.