Browse Articles
Discover research articles across all indexed journals
Frequency-constrained robust unit commitment via physics-guided piecewise-linear nadir surrogates and adaptive virtual inertia
Long-term dynamic effect of body mass index on adverse cardiovascular outcomes with targeted maximum likelihood estimation method: result from the KNOW-CKD study
Paternal heat conditioning enhances offspring’s thermal resilience via epigenetic regulation of mir-210a
Backpropagation-free spiking neural networks with the forward–forward algorithm
Adaptive graph signal processing for robust multimodal fusion with dynamic semantic alignment
Abstract In this paper, we introduce an Adaptive Graph Signal Processing with Dynamic Semantic Alignment (AGSP-DSA) framework to perform robust multimodal data fusion across heterogeneous sources, including text, audio, and images. The proposed approach uses a dual-graph construction to learn both intra-model and inter-modal relations, spectral graph filtering to enhance informative signals, and effective node embeddings via Multi-scale Graph Convolutional Networks. In the semantic-aware attention mechanism, each modality may dynamically contribute to the context with respect to contextual relevance. The experimental outcomes on three benchmark datasets, including Carnegie Mellon University Multimodal Opinion Sentiment and Emotion Intensity dataset, Audio-Visual Event dataset, and MultiModal Internet Movie Database dataset, show that Adaptive Graph Signal Processing with Dynamic Semantic Alignment performs as the state of the art. More precisely, it achieves 95.3% accuracy, 93.6% F1 (Harmonic Mean of Precision and Recall) score, and 92.4% mean average precision on the Carnegie Mellon University Multimodal Opinion Sentiment and Emotion Intensity dataset, improving the MultiModal Graph Neural Network by 2.6% in accuracy. It gets 93.4% accuracy and 91.1% F1 score on Audio-Visual Event dataset, and 91.8% accuracy and 88.6% F1 score on MultiModal Internet Movie Database dataset, which demonstrates good generalization and robustness in the missing modality setting. These findings verify the efficiency of the proposed AGSP-DSA in promoting multimodal learning in sentiment analysis, event recognition, and multimedia classification.
Mainstreaming traditional varieties and on-farm conservation of crop diversity for sustainable finger millet cultivation in Odisha, India
Identification and verification of the key genes, CCR1 and EGR2, in diabetes-associated lipophagy
Abstract Diabetes remains a significant global health challenge, marked by increasing incidence and a complex pathophysiological mechanism involving dysregulated lipid metabolism, impaired autophagy, and chronic inflammatory responses. Lipophagy, an autophagic process that involves the targeting of lipid droplets, is crucial for metabolic homeostasis. Therefore, investigating lipophagy-associated molecules may facilitate the discovery of novel biomarkers and potential therapeutic targets for diabetes. In this study, two GEO datasets, GSE33440 and GSE9006, were combined to identify differentially expressed genes (DEGs) linked to diabetes. By integrating weighted gene coexpression network analysis (WGCNA) with machine learning algorithms, this study identified EGR2 and CCR1 as key hub genes related to lipophagy. The results of the rank sum test revealed a strong positive correlation between these two genes, both of which were significantly upregulated in diabetic samples. Functional analyses, such as gene set enrichment analysis (GSEA), gene ontology (GO) enrichment, and protein‒protein interaction (PPI) network analysis, were used to validate their coherence. Diagnostic models and receiver operating characteristic (ROC) curve analysis further underscore the potential of CCR1 and EGR2 as biomarkers. Importantly, experimental validation demonstrated that the expressions of both genes were significantly elevated in the serum of diabetic patients and in the liver tissues of BKS-db diabetic mice. Notably, compared with control mice, CCR1 knockout mice ( CCR1 -/-) exhibited improved glucose homeostasis under high-fat diet conditions. Collectively, these findings suggest that EGR2 and CCR1 may be potential biomarkers associated with lipophagy in diabetes.
A novel double defected ground structures and parasitic patches for enhanced MIMO antenna performance
Effects of short-term application of organic manure on the growth of forage maize (Zea mays L. cv. Kwangpyeongok) and soil bacterial communities
Thioredoxin protects against diabetic hearing loss by regulating TOMM22 mediated mitochondrial autophagy in hair cells and inhibiting microglial M1 polarization
Abstract Hearing loss is a prevalent yet mechanistically unclear complication of diabetes. This research aims to systematically investigate the role of thioredoxin (Trx) in diabetic hearing loss and elucidate its underlying molecular mechanisms through clinical, bioinformatics, and experimental analyses. Clinically, elevated serum Trx levels were associated with decreased otoacoustic emission (OAE) parameters, suggesting its potential as a serum biomarker. Bioinformatics analysis revealed that TOMM22 is critical for mitophagy. In vivo, Trx overexpression effectively mitigated the hyperactivation of apoptosis, mitophagy, and M1 microglial activation in diabetic mouse cochlear tissue. An in vitro advanced glycation end-product (AGE)-induced HEI-OC1 cell model demonstrated that Trx1 overexpression inhibited PINK1/Parkin-mediated mitophagy, increased the mitochondrial membrane potential, and reduced apoptosis by maintaining TOMM22 expression. Conversely, CCCP-mediated inhibition of TOMM22 reversed the protective effects of Trx1. Moreover, Trx1 drove microglial polarization towards the anti-inflammatory M2 phenotype. Furthermore, coculture of microglia conditioned medium with HEI-OC1 cells confirmed that Trx1 indirectly protects auditory cells. Thus, Trx protects hearing via dual mechanisms: cell autonomously by regulating the TOMM22/PINK1 axis to maintain mitochondrial homeostasis and noncell autonomously by inducing microglial M2 polarization to improve the auditory microenvironment. These findings provide novel insights into diabetic hearing loss pathogenesis and identify Trx as a potential therapeutic target.
Performance analysis of network automation techniques for dense IP networks
Abstract Network automation is an emerging technology which gained a lot of traction over the past few years. NA can be used at multiple levels, such as topology creation, configuration generation, and testing. It can save a great amount of time and effort compared to conventional ways. Here, the automated topology creation tools EVE-NG, Pllama, and Container LAB are evaluated. For configuration generation, Nokia’s Komodo was tested and compared with an automated configuration using python and manually using an Excel sheet. Finally, various test cases were executed in CLASSIC-CLI, MD-CLI, and NETCONF to evaluate the automation performance on the testing level. The results showed that automation greatly reduces the time spent in all the stages of the network deployment mentioned above. In addition, automated topology creation is shown to be 4.5 times faster than in the manual case, configuration generation is about 10% better than the manual case, and test execution time is 11 times faster than manual testing.
A convolutional attention model classifies copy number variants from whole exome sequencing
Abstract Copy number variants are important biomarkers in genetic disease and cancer, yet whole-exome CNV callers often rely on read-depth heuristics that capture limited positional or chromosomal context and generalize poorly across platforms. We present a dual-input convolutional neural network with attention that ingests normalized read depth, genomic coordinates, and chromosome identity. The model was pretrained on ECOLE-labeled 1000 Genomes data and fine-tuned on seven expert-annotated samples. On a held-out test set, the method achieved macro F1 = 0.83 and macro PR-AUC = 0.93. In additional contextual analyses reported in the Supplementary Material, CNN-Att exhibits a sensitivity–precision trade-off consistent with established WES CNV callers. Cross-platform evaluations on HiSeq 4000, NovaSeq 6000, MGISEQ 2000, and BGISEQ 500 yielded overall F1 up to 0.96. Fine-tuning increased deletion and duplication recall at the cost of a higher false positive rate, reflecting an explicit trade-off between sensitivity and precision. The architecture masks padded depth tokens and uses attention to highlight weak depth signals that are characteristic of small or noisy events. These results indicate strong sensitivity and robust performance across sequencing technologies, supporting use in clinical triage, multi-site genomics, and large-scale screening.
Identification and classification of repeated whistle types from free-ranging rough-toothed dolphins (Steno bredanensis)
Abstract Acoustic communication is vital for marine mammals. Many delphinids rely on frequency-modulated tonal signals – known as whistles – used during social communication. Among these, individually distinctive “signature whistles” have been extensively studied in bottlenose dolphins ( Tursiops truncatus ) over the past 60 years. These stereotyped whistles convey individual identity and motivational information. Recent studies have identified similar stereotyped whistles in other delphinids, including a rehabilitated rough-toothed dolphin ( Steno bredanensis ). However, the function, occurrence, time–frequency characteristics, and degree of stereotypy of whistles in free-ranging individuals of this species remain unexplored. This study analysed acoustic recordings from three encounters with rough-toothed dolphins in summer 2023 off Madeira Island (Eastern North Atlantic). Of 4928 whistles analysed, 1015 were identified as repeated whistles and classified into 25 categories based on frequency modulation contours using visual inspection. Visual classification was verified using an inter-observer reliability test and compared to results from an unsupervised neural network (ARTwarp), employing both sequential and global analyses. The results revealed repeated contour types, with the degree of stereotypy varying across categories and generally lower than that reported for other well-studied delphinid species. Like other delphinids, rough-toothed dolphins exhibit a fission–fusion social structure, where repeated signals may help maintain group cohesion, potentially functioning as individual- or group-level calls. Future research incorporating detailed contextual and behavioral information will be crucial to investigate the function, occurrence and social relevance of repeated calls in rough-toothed dolphin whistle repertoires.
Visualization and simulation of full-scale point-neuron circuits via the Neural Circuit Visualizer web platform
AI-assisted reliability-based design framework for tunnel concrete linings in weak rocks
Abstract Reliability-based design (RBD) of tunnel concrete linings in weak rocks is challenging due to the uncertainty of geomechanical parameters and the ground–support interaction. This study introduces a transparent and efficient framework that integrates three key components: (1) implicit ground–support equilibrium analysis based on the Convergence–Confinement Method (CCM), (2) probabilistic evaluation of the safety factor ( $$\:SF$$ ), reliability index ( $$\:\beta\:$$ ), and failure probability ( $$\:{P}_{f}$$ ) and (3) a data-driven surrogate model based on artificial intelligence for rapid parametric analysis. The uncertainties of the rock mass and concrete linings are treated through Monte Carlo sampling, and the outcomes are benchmarked against the First-Order Reliability Method (FORM) to verify accuracy and potential bias. The results are transformed into decision-making charts linking the thickness and compressive strength of concrete linings to target $$\:\beta\:/{P}_{f}$$ levels, thus providing risk-consistent objectives instead of a fixed safety factor criterion. Two types of linings are considered: conventional concrete (with an average uniaxial compressive strength of 20 MPa) and fiber-reinforced reactive powder concrete (FRPC, with an average uniaxial compressive strength of 65 MPa, developed by the research team). The findings demonstrate that increasing thickness and material quality significantly reduce $$\:{P}_{f}$$ , achieving the reliability thresholds required for final tunnel support (2E-5). This effect is more critical at smaller thicknesses and with lower-quality materials. The use of FRPC also leads to a considerable reduction in $$\:{P}_{f}$$ at intermediate thicknesses, making it an efficient option when construction constraints limit thickness increase. The surrogate model successfully reproduces probabilistic trends with high consistency and slight conservatism. Ultimately, the integrated CCM–reliability–artificial intelligence framework bridges the gap between deterministic design and risk management, delivering economical and resilient tunnel lining designs in weak rocks.
Comparing novel backward hydrological models for watershed-scale precipitation estimation: an evaluation of inverted PDM and Kirchner-hybrid structures
Correction: Achievable directive antenna gain measurements and modeling for 60 GHz indoor links
Silica based ZnFe2O4 nanocomposite as a novel photocatalyst for basic fuchsin dye degradation
Abstract Growing environmental problems and the threat of an energy crisis have created an urgent need for affordable and efficient photocatalysts that can work under UV light to remove pollutants. In this work, Zinc ferrite (ZnFe 2 O 4 ) nanoparticles were prepared by a chemical co-precipitation method, while SiO 2 nanoparticles were produced through dry mechanical milling. These materials were then combined in a 1:1 molar ratio using the same dry mechanical method to form a doped ZnFe 2 O 4 /SiO 2 nanocomposite. The structure and composition of the composite were analyzed using high-resolution transmission electron microscopy, scanning electron microscopy, and energy-dispersive X-ray spectroscopy. The photocatalytic and adsorption performance of the ZnFe 2 O 4 /SiO 2 nanocomposite was tested under different conditions, including pH, initial dye concentration, and amount of nanocomposite used. The results showed that 0.01 g of the composite removed 95% of basic fuchsin dye at pH 11 after 150 min. The kinetic analysis revealed that the degradation process followed a pseudo-first-order model. Overall, the ZnFe 2 O 4 /SiO 2 nanocomposite showed excellent potential as an effective and practical material for environmental cleanup and industrial wastewater treatment.