Search

Early identification of high-risk trauma patients in prehospital settings is essential for effective triage and improved survival. Here the authors show that a real-time ensemble AI model accurately predicts emergency department mortality using only prehospital data, outperforming traditional triage tools.

Risk stratification in non-small cell lung cancer (NSCLC) remains challenging. By combining multiplex immunofluorescence, H&E histology, and AI, the study identifies spatial “cell-niche” patterns that enhance survival prediction beyond UICC8 staging. These patterns reclassify many stage I patients as high risk, revealing potentially undertreated cases and establishing spatial tumor microenvironment features as clinically actionable biomarkers.

Global Burden of Disease analysis found that in 2023, suboptimal diet was estimated to be responsible for 4 million deaths and 97 million morbidity burdens from ischemic heart disease.

In this Comment, Na Sun explores how artificial intelligence-driven methodologies are poised to transform systems biology, particularly in the realm of tissue modelling, and outlines the key challenges that must be overcome to enable the development of truly predictive biological systems.

Technical advancements have significantly improved early diagnosis of cervical cancer, but accurate diagnosis is still difficult due to various practical factors. Here, the authors develop an artificial intelligence assistive diagnostic solution to improve cervical liquid-based thin-layer cell smear diagnosis according to clinical TBS criteria in a large multicenter study.

This study identifies an FDA-approved leukemia drug, Homoharringtonine, that can eliminate aged fat cells, improve metabolic health, reduce inflammation and extend lifespan in mice, revealing a potential approach to treat age- and obesity-related metabolic disease.

Traditional concrete testing can be labour-intensive and limited in accuracy, consistency, and real-time applicability. The study uses piezoelectric sensors and deep learning for real-time monitoring of concrete strength, interpreting signals to achieve accurate predictions. This is further validated in highway projects.

A chemist-intuited atomic robotic probe is developed that enables autonomous site-selective manipulation of magnetic nanographenes with atomic precision and aids in reaction mechanism elucidation through the incorporation of learned knowledge and artificial intelligence, leading to the intelligent synthesis of these materials.

A multimodal model for the stratification of breast cancer risk based on clinical metadata, mammography and trimodal ultrasound images performed as well as or better than radiologists at tumour classification and at differential diagnosis.

Organohalides play roles in marine microbial ecology and biogeochemical cycles. Here, the authors analyse marine metagenomic datasets and construct a database of genes and microorganisms involved in microbial halogenation and dehalogenation of organic compounds in the oceans.

Sustainable electrification is essential for addressing climate change and leveraging artificial intelligence (AI). Electric grids have a fundamental role in decarbonizing the economy and enabling AI. Here we propose a comprehensive research agenda that integrates innovations in material discovery, computer architecture, smart grids and system theory to accelerate efficient, large-scale, low-carbon electrification.

The chromatin remodeler BRAHMA (BRM) is found as a repressor of plant adaptation to acid stress by inhibiting the STOP1-NRT1.1 signaling module. Under low pH, BRM degradation activates nitrate uptake and alkalinizes the rhizosphere.

In this work, functional antibodies against the cancer-related GPCR NTS1 were generated by immunizing chickens with in silico-designed 3D peptides, offering a strategy for GPCR targeting with potential diagnostic and therapeutic applications.

Here they show that NONO collaborates with HOXA1 to modulate gene expression during early cardiomyocyte differentiation. This interaction facilitates activation of the Wnt signaling pathway, establishing the NONO-HOXA1-Wnt axis as a key mechanism in cardiac development.

Developed on cytology images of hydrothorax and ascites from 57,220 cases at four hospitals, a deep-learning model shows high accuracy in tumor origin prediction and presents prognostic value when patient treatment is consistent with the cancer origin predicted by the model.

Sensor failures and limited resolution challenge many complex systems. Here, authors develop a multimodal AI method to generate super-resolution of a sensor using other available sensors in the system, revealing hidden dynamics in fusion plasmas and enabling cost-effective, high-resolution diagnostics.

China’s dramatic demographic shift toward population aging raises challenges at the individual, familial and societal levels. Fang et al. review these challenges and emerging policies designed to promote healthy longevity in China.

Hashimoto’s thyroiditis (HT) is the main cause of hypothyroidism. Here the authors develop a deep learning model for diagnosis of HT on a large multi-site dataset including image and video data.

Immune check point therapies have not been successful for prostate cancers, but the underlying mechanisms are unclear. Here the authors show that a transcription factor, YY1, may promote the SUMOylation and stability of HIF-1α to suppress anti-tumor immunity, while targeting YY1 helps improve tumor control, in mouse prostate cancer models.

Song et al. developed a computational tool to profile binding pairs between tumor-derived antigens and antibodies from high-throughput B cell receptor sequencing data, predicting response to immune-checkpoint inhibitor therapy and risk of adverse events.

Liu et al. developed AI-driven, wearable olfactory interfaces with odor generators for realizing latency-free mixed-reality and fast olfaction recovery. This technology enables personalized olfactory feedback and enhanced virtual environments for various applications including education, and clinical treatment.

A virtual process game to benchmark the performance of humans and computers for the fabrication of semiconductors leads to a strategy combining human expert design with optimization algorithms to improve semiconductor process development.

Soil carbon content is positively related with plant diversity in global grasslands, and this relationship is particularly strong in warm and arid climates. Plant diversity is related to soil carbon via the quality of organic matter.

A hybrid cathode interfacial layer (AZnO-F3N) is developed, delivering 21.0% efficiency along with excellent stability, mechanical robustness and broad versatility, highlighting its potential to advance organic solar cells.

A near-infrared fluorescent calcium indicator can be combined with other optogenetic tools in vivo without spectral crosstalk.

Spermine is a polyamine dysregulated in different cancers. This group shows that Serpinb9 prevents spermine synthase degradation, raising spermine levels and impairing the tumor immune microenvironment, and identifying Serpinb9 as a potential therapeutic target for pancreatic cancer immunotherapy.

Hypercompact type V CRISPR-Cas12n nucleases target DNA sequences containing rare A-rich PAMs. Here, authors reveal the cryo-EM structure of RdCas12n, detailing its RNA-guided DNA targeting and evolutionary origins, and engineer it for effective genome editing in human cells.

The authors measured the orientation tuning of ∼28,000 thalamic boutons and ∼4,000 neurons in layers 1–5 of awake mouse V1. With adaptive optics allowing accurate measurement of deep tissue activity, around half of the boutons in layer 4 were found to carry orientation and direction information.