Gas sensing material screening faces challenges due to costly trial-and-error methods and the complexity of multi-parameter interactions. To address this, this study integrates first-principles calculations with machine learning (ML) for rapid gas sensitivity prediction. Using black phosphorus (BP) as a model, this study analyzes adsorption-induced changes in electronic and structural properties across 21 gases. Key descriptors derived from first-principles calculations train six ML models, with the Extra Trees (ET) model achieving 96% accuracy and top F1-scores in validation. SHAP analysis identifies adsorption energy, p-orbital center, valence/conduction band edges, and Fermi level as dominant descriptors. Morover, a lightweight Python-based system enables real-time response prediction using these five features, demonstrating strong potential for guided sensor design.

Ultrasonic-assisted hot pressing (UAHP) has shown significant potential in enhancing both densification and mechanical performance of metallic materials. However, the poor high-temperature stability of the ultrasonic system severely limits its application in the fabrication of high-melting-point materials. Currently, UAHP has yet been applied to difficult-to-sinter ceramic materials, such as monolithic boron carbide (B₄C). Filling this research gap is imperative since UAHP opens a new avenue for the preparation of difficult-to-sinter ceramics.

The electrochemical oxidation of glycerol (GOR) is gaining traction as a sustainable method to convert biodiesel byproducts into valuable chemicals and fuels, aligning with global demands for renewable energy and green production. Recent advances in catalyst design, reaction mechanisms, and system integration are driving progress, though challenges in selectivity, stability, and scalability remain pivotal for industrial adoption. Researchers are tuning both noble and non-noble metal catalysts—through methods such as facet engineering and single-atom doping—to selectively steer reactions toward high-value multi-carbon products. Furthermore, coupling GOR with cathodic processes like hydrogen evolution or CO₂ reduction offers a path to lower energy use and co-produce clean fuels. Key hurdles, including mass transfer limits and feedstock compatibility, still need addressing. Proposed solutions range from advanced electrode assemblies to integrated techno-economic assessments. Moving forward, a system-level approach that balances technical performance with economic viability will be essential to accelerate GOR technology toward real-world application.

Exosomes facilitate cell-to-cell communication and are involved in key biological processes. Understanding the mechanisms regulating exosome production could offer new therapeutic insights for various diseases. Here, Prof. Zhong’s team demonstrates that exosome secretion is significantly inhibited when glucose is replaced with galactose as the primary carbon source in the culture medium. This glycometabolic regulation of exosome secretion is dependent on the cellular hexosamine biosynthetic pathway (HBP). Inhibition of HBP via gene knockdown, pharmacological blockade, or metabolite deprivation markedly suppresses exosome secretion. Mechanistically, HBP regulates multivesicular body (MVB) outward trafficking and its fusion with the plasma membrane via synaptosomal-associated protein 25 (SNAP25). O-GlcNAcylation of SNAP25 promotes soluble N-ethylmaleimide-sensitive factor attachment protein receptor (SNARE) complex assembly, thereby facilitating exosome release. In summary, these findings reveal a critical role of HBP and protein O-GlcNAcylation in exosome secretion, which may provide new therapeutic targets for exosome-associated diseases, including cancer and inflammatory disorders.

In the field of polyoxometalate chemistry, organophosphonate covalently modified polyoxometalates have recently emerged as a promising frontier. These hybrid materials not only broaden the structural diversity of conventional polyoxometalate derivatives and address the inherent stability limitations of polyoxometalates, but also allow for the design of improved properties tailored to diverse applications. This review provides a comprehensive summary of recent advances in organophosphonate covalently modified polyoxometalates research, with a particular focus on their structural features, functional properties, and prospective research directions.

Digital twin (DT) technology is emerging as a core solution for future marine development and intelligent ocean management. The review systematically reviews digital twin applications in the marine field, clarifies its concept, proposes a five-layer framework, and summarizes key technologies, including sensing, data management, modeling, simulation, and monitoring. It highlights DT’s ability to synchronize physical marine systems with virtual models in real time, enabling simulation, prediction, optimization, and decision-making. The authors further outline challenges and development prospects, showing how DT can support deep-sea resource exploitation, offshore wind energy, marine engineering, vessel autonomy, environmental monitoring, and system reliability assessment.

Polyoxometalates are promising inorganic drugs with antiviral activity; however, they pose a risk to humans because of their potential accumulation in the body. Polyoxometalates encapsulated with berberine from a traditional Chinese herb may exhibit lower cytotoxicity. In this study, the antiviral effects of four berberine-based organic–polyoxometalate hybrids (BR-POMs) on BHK-21 and PK-15 cells were evaluated in vitro using encephalomyocarditis virus (EMCV) or pseudorabies virus (PRV) models. The collected cells were used for quantitative polymerase chain reaction analysis. The supernatants were collected to quantify the viral loads using a TCID50 assay in vitro. EC50 and CC50 were determined through dose–response experiments, and the EC50/CC50 ratio was used as a selectivity index to measure the antiviral activity. The results demonstrate that all BR-POMs exhibited certain antiviral activity. The BR-POMs did not exert toxicity against the EMCV- or PRV-infected cells at the tested concentration (CC50 > 40 μM). Notably, BR-EuSiW (EC50 15.07 μM, CC50 651.2 µM, SI 43.21) exerted antiviral effects by acting on the virus at its biosynthesis stage, thereby inhibiting virus proliferation in a dose-dependent manner. This study demonstrates that organic–polyoxometalate hybrids represent a new strategy for developing antivirals against EMCV.

Multimodal large language models have shown powerful abilities to understand and reason across text and images, but their massive size and computational cost limit real-world deployment. This research systematically examines how multimodal models can be made more efficient without severely sacrificing performance. By analyzing lightweight architectures, visual token compression strategies, efficient training methods, and compact language backbones, the study maps out the technical pathways that reduce memory demand and inference cost. The work highlights how efficiency-focused design enables multimodal intelligence to move beyond cloud-based systems toward broader, more accessible applications, including mobile devices and edge computing environments.