Browse Articles
Discover research articles across all indexed journals
Identification of high-yielding and stable genotypes of barley in the cold climate of Iran using AMMI and GGE biplot models
Sialic acid-anchored haemagglutinin stalk neutralizing antibody M-SiaB enhances protection against highly pathogenic influenza H5N1/Texas/2024
A cloud-edge adaptive lightweight network with dynamic inference enables real-time ultrasound image segmentation
Abstract Ultrasound image segmentation plays an important role in clinical diagnosis, but existing deep learning methods struggle to balance accuracy and efficiency, limiting practical deployment in resource-constrained medical scenarios. This study proposes a Cloud-Edge Adaptive Lightweight Network (CEA-Net) with dynamic inference, implementing parameter compression through a lightweight encoder constructed with depthwise separable convolutions and inverted residual structures, introducing serial channel-spatial attention mechanisms to enhance lesion feature representation, and designing intelligent routing algorithms based on sample complexity perception to dynamically select edge fast inference or cloud enhanced processing paths according to ultrasound image complexity. Experimental validation on the BUSI breast ultrasound dataset and TN3K thyroid nodule dataset demonstrates that CEA-Net significantly reduces model parameters and inference latency while maintaining high segmentation accuracy, achieves near real-time inference capability on edge devices, and reaches a favourable balance between accuracy and speed through intelligent task allocation in cloud-edge collaborative mode. This study provides a feasible technical solution for practical deployment of ultrasound image segmentation technology in resource-constrained scenarios including primary healthcare institutions, telemedicine, and mobile diagnosis, advancing the practical application of medical artificial intelligence technology from cloud to edge.
The chemistry of habitable oceans in the Solar system
A two-stage deep learning approach for facial emotion recognition in RAVDESS videos
Abstract Video-based emotion recognition is an important topic in affective computing, with applications in human–computer interaction, mental health, and multimedia systems. In this work, we propose a two-stage deep learning approach that extracts spatial features from individual frames and leverages temporal consistency across video sequences for facial expression recognition using the RAVDESS dataset. First, videos are split into frames, and a fine-tuned VGG16 CNN extracts discriminative spatial features from each frame. Second, these features are aggregated into sequences and the frame-level predictions are aggregated at the video level using a majority voting strategy to ensure temporal consistency. Our experiments show that the proposed method achieves 93.6% accuracy at the frame level and 98.1% at the video level, outperforming baseline models and remaining competitive with recent state-of-the-art approaches. Temporal aggregation helps reduce misclassifications of subtle emotions, while fine-tuning improves feature extraction. The approach is computationally efficient and provides a solid foundation for future research in multimodal emotion recognition and advanced video-level aggregation.
Tissue-adhesive hydrogel optical fiber for peripheral optogenetic neuromodulation
Abstract While hydrogel optical fibers hold promises for visceral peripheral optogenetics, their utility is limited by poor tissue adhesion, unstable light delivery, and micromisplacement under physiological motion, leading to off-target illumination. To address these challenges, we developed tissue-adhesive hydrogel optical fibers (TAHOFs) integrating a poly(HEMA) light-guiding core ( n core = 1.429 ± 0.004) with a bioadhesive cladding ( n cladding = 1.343 ± 0.002), achieving dual functionality through refractive index contrast (Δ n = 0.086) and robust tissue integration (11.5 ± 1.8 kPa). This architecture enables efficient optical confinement with low propagation loss (0.534 ± 0.092 dB/cm) while maintaining spatial targeting fidelity under 30% tensile strain. Pancreatic implantation in freely moving ChAT-ChR2 mice demonstrated precise vagal fiber activation, effectively triggering insulin secretion through mechanically and optically stable light delivery. Integrated with subcutaneous continuous glucose monitoring, TAHOFs enabled 3-day glycemic control in diabetic models, showing stable blood glucose reduction and real-time regulation. Chronic implantation demonstrated that TAHOF supports stable in vivo adhesion on the pancreas while maintaining optogenetic function for up to 14 days. The TAHOF uniting high optical efficiency, mechanical compliance, and biological integration, offering an application-specific design strategy for optogenetic neuromodulation in moving animals, particularly mechanically dynamic and anatomically complex organs.
Dynamic reconfiguration of subunits from the hippocampal-amygdala complex indicate patterns of psychosis vulnerability in 22q11.2 deletion syndrome
Abstract The hippocampus and amygdala form a tightly integrated circuit supporting memory-emotion integration and stress regulation, critically implicated in psychosis development. However, the maturation of the hippocampus-amygdala (HIP-AMY) complex as a functional unit remains poorly understood. 22q11.2 deletion syndrome (22q11DS), with increased risk of schizophrenia, provides a valuable model to study its neurodevelopmental trajectories. A prospective longitudinal cohort of 133 individuals with 22q11DS and 150 healthy controls (HC) underwent neuroimaging and psychiatric assessments approximately every three years. Resting-state fMRI data were analysed using a data-driven clustering approach to identify transient states of dynamic functional connectivity linked to the activation of specific ensembles of HIP-AMY nuclei. State occurrences over age were compared between 22q11DS and HC and between 22q11DS individuals with and without positive psychotic symptoms (PPS+ and PPS−). Seven HIP-AMY states were identified, three showing significant and one trend-level age-by-diagnosis interaction. Importantly, the occurrence of a CA3-centromedial amygdala driven state showing co-activation of limbic and thalamo-striatal salience hubs decreased over age in HC, but remained elevated in 22q11DS. This divergence was reflected within 22q11DS, with PPS− showing a decrease and PPS+ a modest increase. The persistence of this pattern over development may provide a dynamic signature of emergence of psychosis.
A pilot translational study of neoadjuvant fulvestrant plus abemaciclib in women with advanced low-grade serous carcinoma
Bioprospecting spore-forming bacteria: bioactivity-guided fractionation and profiling of metabolite-enriched fractions from Lysinibacillus fusiformis IWSA
The structural mechanism of HIV-2 Vif antagonism of human APOBEC3H
Abstract Human APOBEC3 (A3) proteins restrict retrovirus infection by inducing hypermutations in viral cDNA. To counteract this restriction, lentiviruses, such as HIV-1 and HIV-2 encode the viral infectivity factor (Vif), which hijacks a host Cullin-RING E3 ubiquitin ligase complex to target A3 proteins for proteasomal degradation. Here, we present the cryo-EM structure of HIV-2 Vif in complex with human A3H and CBFβ. The structure reveals that A3H forms a dimer mediated by dsRNA where each A3H monomer directly interacts with an HIV-2 Vif and the host protein CBFβ. Both HIV-2 Vif-A3H and CBFβ-A3H interfaces are critical for A3H degradation. Notably, however, the HIV-2 Vif-A3H interface is entirely distinct from the previously determined cryo-EM structure of the HIV-1 Vif and A3H complex. These findings suggest that HIV-1 and HIV-2 Vif, which are the result of distinct cross-species transmissions from species with different A3H characteristics, have followed separate evolutionary trajectories to counteract human A3H.
Maternal undernutrition during gestation induces enduring genome-wide DNA methylation alterations in the skeletal muscle of postnatal beef cattle
Abstract Maternal nutrition exerts long-term effects on offspring development, potentially mediated by epigenetic mechanisms. This study aimed to characterize the genome-wide DNA methylation profile of skeletal muscle in postnatal cattle and to determine the long-term impacts of maternal undernutrition on DNA methylation in offspring muscle. Wagyu cows were assigned to either nutritional-adequate control (CNT; n = 4, 120% of requirements) or nutritional-restricted (NR; n = 4, 60% of requirements) group from day 35 of gestation until parturition. After birth, all offspring received identical diets, and longissimus thoracis muscle (LM) biopsies were collected at 300 days of age for DNA methylation analysis using whole-genome bisulfite sequencing. Irrespective of maternal diet, DNA methylation levels in offspring muscle gradually decreased across the CpG from the upstream region toward the transcription start site, reached their lowest level at the transcription start site, and increased throughout the gene body. Compared with the CNT group, NR offspring LM exhibited 7076 hypomethylated and 6104 hypermethylated regions (|methylation difference| > 20%, Q < 0.05). Among genomic features, promoter and 5′ untranslated regions exhibited the greatest susceptibility to methylation changes, with 0.96% and 1.09% of these regions being hypomethylated in NR offspring LM relative to CNT. Genes containing differentially methylated regions in distal upstream (1–5-kb upstream from transcription start site) regions or promoters were associated with fundamental biological processes such as gene expression regulation, protein function, cell and tissue development, cytoskeletal and contractile organization, and neurodevelopment ( P < 0.05). An overlap-based integrative analysis of DNA methylation and gene expression data identified seven candidate epigenetically regulated genes, including neuronal precursor cell-expressed developmentally downregulated 4 (Entrez Gene ID: 507781) and solute carrier family 30 (zinc transporter) member 1 (Entrez Gene ID: 522265). Although only seven candidate genes were identified through integrative analysis, more than 13,000 differentially methylated regions were maintained in offspring muscle. These findings suggest that maternal undernutrition induces changes in DNA methylation patterns during the fetal stage that persist postnatally, and may contribute to long-term effects on muscle metabolism, growth efficiency, and meat quality.
Underwater Suit-Wearing Cyborg Insect Capable of Hours-Long Diving and Terra-Aqua Travel
Abstract The fundamental operational range of cyborg insects, which are hybrid robots that combine a living insect with an electronic controller, is inherently restricted to the host’s natural environment. To extend their operational range, we developed a wearable diving suit for terrestrial insects. The suit integrates a miniaturised oxygen generation module with a flexible waterproof shell, enabling continuous oxygen supply and isolation from surrounding water. By fitting a cockroach, which is a terrestrial species, into this diving suit, we allowed it to survive and operate in oxygen-deprived environments such as underwater, transforming it into an amphibious cyborg robot capable of operation across land and water. The suit sustained respiration and locomotion for up to 3 h underwater, establishing amphibious cyborg insects that combine biological adaptability with engineered protection for prolonged exploration in extreme, confined environments.
Plasmonic graphene oxide aptasensor empowered for ultra‑sensitive and rapid detection of aflatoxin M1 in dairy product
Abstract Aflatoxins are toxic metabolites and potent carcinogens that pose a threat to both humans and animals even at trace levels. In this study, we developed a robust plasmonic graphene oxide-based aptasensor for rapid and specific detection of Aflatoxin M1(AFM1) in dairy products. The Graphene Oxide–Chitosan (GOCS) nanocomposite was applied onto an Au thin layer (gold chip) and functionalized with aptamers. The Finite-Difference Time-Domain (FDTD) technique was employed to simulate the evanescent field and surface plasmon resonance (SPR) behavior of the sensor, enabling optimization of its performance. Additionally, X-ray Diffraction (XRD) and Field Emission Scanning Electron Microscopy (FE-SEM) analyses were conducted to examine the structure of GOCS/Au and the properties of the SPR sensor. The aptasensor successfully detected AFM1 in milk with a detection limit of 0.005 ng/mL and a linear range from 0.005 to 50 ng/mL. Given both the sensor’s linear range and its low detection limit, it readily meets and surpasses the regulatory requirements of the European Union and the United States for permissible AFM1 levels in milk. The performance of the sensor was validated under various pH conditions and further compared to the High-Performance Liquid Chromatography (HPLC) method in the presence of other mycotoxins. The results demonstrated the high sensitivity and selectivity of the sensor, highlighting its strong potential for reliable AFM1 detection in diverse food samples.
Impact of molecular multimodality on neural network models for prediction tasks related to drug discovery
Social support and self-efficacy mediate the link between health literacy and fear of disease progression in Chinese lung cancer patients
Abstract Fear of disease progression (FOP) has emerged as one of the most prevalent and distressing unmet needs among lung cancer patients in recent years. This study aimed to explore the associations between health literacy (HL), social support (SS), self-efficacy (SE) and FOP, and to examine the mediating roles of SS and SE in the relationship between HL and FOP in lung cancer patient population. A cross-sectional design with convenience sampling was employed to survey cancer inpatients in two general hospitals in Chongqing and Chengdu, two major cities in southwestern China. Data were collected via validated scales for HL, SE, SS and FOP. A structural equation model (SEM) was then constructed, P ≤ 0.05 was considered statistically significant. The mean FOP score FOP was 31.20 ± 9.28; 162 patients (43.09%) exceeded cutoff for psychological dysfunction (≥ 34). Correlation analysis revealed that FOP was significantly negative correlated with all variables except for objective support (OS) and subjective support (Subs) ( p < 0.01). The SEM of the FOP exhibited good fit (GFI = 0.949, NFI = 0.964, CFI = 0.978, IFI = 0.978, AGFI = 0.918, RMSEA = 0.064). The model indicated that HL had the strongest correlation with FOP (β = -0.300, p < 0.05), followed by SE (β =−0.281, p < 0.05). Notably, SE mediated the association between HL and FOP (contribution effect: 40.67%). SS positively influence SE (β = 0.131, p < 0.05) but had no statistically significant effect on FOP ( p > 0.05). HL exerted the strongest protective effect on FOP, both directly and indirectly via enhanced SE. Utilization of support Subs (US) was clinically more significant than OS and Subs. Clinical interventions should prioritize improving health literacy through targeted education, psychological support, and coping skills training to strengthen patients’ use of available support and foster psychological adaptation.
Cobalt-catalyzed asymmetric three-component reductive addition of alkynes with alkyl halides and imines
Correlation Study between behavioral inhibition/activation system, negative emotions and mindfulness in patients after hip fracture surgery
HSV-1 origin binding protein UL9 forms intermolecular DNA tethers and intramolecular DNA loops
Abstract Herpesviruses are ubiquitous human pathogens, causing mild to severe symptoms ranging from cold sores to nasopharyngeal carcinoma. Even though replication of the dsDNA genome has been studied for decades, we still lack a complete molecular understanding of its mechanism. It has been previously proposed, but never shown directly, that the HSV-1 origin-binding protein UL9 interacts with two closely spaced sites within the oriS origin sequence, thereby mediating origin looping, which in turn facilitates replication initiation. Here, we use an array of single-molecule approaches to test this long-standing hypothesis directly. Surprisingly, the data show that UL9 does not efficiently loop oriS. However, we demonstrate that UL9 can form large DNA loops at non-origin sequences very efficiently, as well as tether two oriS DNA molecules intermolecularly. Contrary to the origin-bending hypothesis, our findings indicate that UL9 does not primarily loop oriS DNA but rather may play an alternative role in replication initiation, such as tethering two separate molecules to facilitate recombination.
Negotiating knowledge: The role of network hedging in the production of high-impact science
Extant research shows that knowledge networks with a greater diversity of participants and richer in structural holes are more conducive to advancing science and innovation. However, research on network structure and network composition tends to overlook how actors might best utilize their connections. In this study, we seek to unpack the variation in how individuals leverage the opportunities afforded by their networks, focusing on the scientific knowledge production process. Given the uncertainty faced by scientists during the different stages of the research process, we argue that network hedging – consulting multiple individuals for the same resource need – rather than network compartmentalization – turning to a particular network contact for a specific resource need – increases the likelihood of research findings with greater scientific impact. By analyzing granular data on the network mobilization decisions and scientific outputs for a sample of biomedical scientists, we find support for our prediction that network hedging is positively associated with the production of high-impact scientific output, and that this effect is manifest beyond the benefits of being embedded in more diverse or sparser networks.