Glycolipid metabolism reprogramming and transcriptional splicing of ARID4B genes during the response to summer seawater warming in Patinopecten yessoensis.

Linguistic abnormalities in schizophrenia (SCZ) span morphological, syntactic, semantic, and discourse levels. Converging cross-linguistic evidence suggests that SCZ may involve semantic narrowing alongside reduced syntactic differentiation, yet how these changes co-occur across linguistic domains and whether they represent core, task-general disturbances remains unclear. We applied a multilevel NLP framework to a large Japanese dataset to identify structurally related linguistic markers of SCZ across elicitation contexts. Speech from 104 patients with SCZ and 101 healthy controls was collected through semi-structured interviews. Transcripts from free conversation, storytelling, and picture description were analyzed using GiNZA, Word2Vec, TF-IDF, and SentenceBERT to extract 76 morphosyntactic, semantic, and discourse features. Factor analysis identified representative features independent of diagnosis, which were tested using generalized estimating equations and validated with bootstrap and permutation procedures. Cross-task stability was examined to determine core linguistic markers. In free conversation, reduced Case-particle (Kakujoshi) and Adverb use and increased Mean Pairwise Word Similarity were strongly associated with SCZ (AUC=0.87, 95% CI: 0.74-0.97). Adverbial, case-particle, and semantic-network measures functioned as cross-task markers. SCZ involves multidimensional language disturbances characterized by a tripartite linguistic phenotype of diminished morphosyntactic explicitness, semantic narrowing, and reduced modification-based contextual modulation in spontaneous discourse. Extending cross-linguistic evidence, our results indicate that lexical-semantic contraction co-occurs with reduced overt marking of argument relations in Japanese, alongside weakened adverbial elaboration and framing - suggesting convergent, largely task-general dimensions of SCZ language pathology, most evident in free conversation.

Multi-omics characterization of glucose-lipid metabolic remodeling associated with Gymnema sylvestre (Retz.) R.Br. ex Sm. treatment in a type 2 diabetic mouse model.

In the increasingly competitive New Energy Vehicle (NEV) market, consumers’ purchase decisions are influenced not only by performance but also by emotional responses evoked by vehicle appearance. However, existing studies lack a robust framework to translate such preferences into concrete design parameters and to resolve the mapping between vague emotional needs and design features. To address this gap, this paper proposes an attractive NEV form design approach integrating the Interval Type-2 Trapezoidal Fuzzy Sets-Kano Model (IT2Tr-FKM) with the Hippopotamus Optimisation Algorithm-eXtreme Gradient Boosting (HO-XGBoost). Based on a three-level evaluation structure constructed using the Evaluation Grid Method (EGM) of Miryoku Engineering, IT2Tr-FKM is used to evaluate and prioritise upper-level Kansei words. Morphological decomposition then links key Kansei words to abstract reasons and concrete design attributes. HO-XGBoost establishes a mapping between Kansei words and representative design features to identify high emotional appeal solutions. Eye-tracking and the Data Information Difference Fluctuation Weighting Method (DIDF) are further applied for objective and subjective evaluation, ultimately selecting the optimal NEV form design. The proposed framework improves design efficiency and user satisfaction.

Accurate online detection or prediction of key quality variables provides critical reference information for optimizing and controlling operating variables in industrial processes. However, frequent fluctuations in raw material properties and environmental conditions often give rise to multiple data distribution modes within the same production process. Moreover, the inherent uncertainties and the energy-material coupling characteristics of industrial processes make it particularly challenging to uncover the underlying topological relationships among process variables. To address these issues, this article proposes a novel jointly shared-specific variational graph attention autoencoder (JSS-VGATE) model for spatial topological feature extraction and key quality variable prediction in multimode industrial processes. Specifically, a variational graph attention autoencoder is first constructed, which combines graph attention mechanisms with the variational inference architecture to adaptively learn the dynamic correlation strengths between adjacent nodes, thereby capturing complex variable interactions. Subsequently, a comprehensive loss function is designed to achieve high-fidelity extraction of representative latent feature distributions. Furthermore, a cross-mode jointly shared-specific learning framework is developed to simultaneously capture global shared features across modalities and preserve local specific features of each modality, while a learnable gated fusion mechanism is introduced to balance modality invariance and heterogeneity, thereby enhancing cross-mode information integration. Finally, the effectiveness and superiority of the proposed JSS-VGATE are validated on two representative real-world industrial datasets compared to other state-of-the-art methods.

Speech-derived acoustic biomarkers for depression: Comprehensive cross-section and longitudinal analyses in different cohorts.

Continual learning aims to enable models to acquire new knowledge while retaining previously learned information. Prompt-based methods have shown remarkable performance in this domain; however, they typically rely on key-value pairing, which can introduce inter-task interference and hinder scalability. To overcome these limitations, we propose a novel approach employing task-specific Prompt-Prototype (ProP), thereby eliminating the need for key-value pairs. In our method, task-specific prompts facilitate more effective feature learning for the current task, while corresponding prototypes capture the representative features of the input. During inference, predictions are generated by binding each task-specific prompt with its associated prototype. Additionally, we introduce regularization constraints during prompt initialization to penalize excessively large values, thereby enhancing stability. Experiments on several widely used datasets demonstrate the effectiveness of the proposed method. In contrast to mainstream prompt-based approaches, our framework removes the dependency on key-value pairs, offering a fresh perspective for future continual learning research.

Diabetic foot infections (DFIs) represent a major public health concern, and methicillin-resistant Staphylococcus aureus (MRSA) is among the most clinically significant pathogens. This study investigated the prevalence of virulence genes (cna and hlg), antimicrobial resistance profiles, and representative genomic features of multidrug-resistant S. aureus isolates recovered from patients with DFIs. A cross-sectional observational study was conducted on 125 patients with diabetic foot infections between January and December 2024. Antimicrobial susceptibility testing was performed using the Kirby-Bauer disk diffusion method and cefoxitin screening according to Clinical and Laboratory Standards Institute (CLSI) guidelines, with S. aureus ATCC 25923 used as the quality-control strain. Vancomycin susceptibility was confirmed by minimum inhibitory concentration (MIC) testing. Polymerase chain reaction (PCR) was used to detect the virulence genes (cna and hlg) and blaOXA-group I genes. Whole-genome sequencing (WGS) was performed on two representative isolates, including one multidrug-resistant (MDR) isolate and one extensively drug-resistant (XDR) isolate, using a de novo sequencing approach to generate draft genome assemblies. Among 125 clinical specimens, bacterial growth was observed in 90 samples (72%), of which 45 isolates (50%) were identified as S. aureus. Among these isolates, 35/45 (77.78%) were classified as MRSA, and 36/45 (80%) were multidrug resistant. The hlg gene was detected in all isolates, whereas the cna gene was identified in 13/45 (28.89%) isolates. No blaOXA-group I genes were detected. Genomic analysis identified multiple resistance-associated genes, including blaZ, tet(38), norA, and vanT, together with CRISPR-Cas elements and plasmid-associated resistance determinants. These findings highlight the high prevalence of multidrug-resistant S. aureus in DFIs and support the importance of continued genomic surveillance of clinically relevant resistant strains.

Peanut (Arachis hypogaea L.) is a prominent legume and oilseed crop cultivated globally for its economic significance and nutritional value. With the growing demand for edible oil and protein, peanut cultivation needs to be expanded into salinized land. However, salt stress in such soils severely limits peanut growth and productivity. The Na⁺/H⁺ antiporter (NHX) gene family plays a crucial role in plant salt tolerance by regulating ion homeostasis. Genome-wide identification of peanut NHX genes and systematic profiling of their expression patterns under salt stress are essential for enhancing peanut salt tolerance through molecular breeding. In this study, we conducted a genome-wide analysis of the NHX gene family in cultivated peanut and identified 73 AhNHX genes. Based on NHX protein sequences from Arabidopsis thaliana, Glycine max, Medicago sativa, Medicago truncatula, and Phaseolus vulgaris, a phylogenetic tree was constructed, and the 73 AhNHX genes were divided into three clades. Furthermore, analysis of gene structure and conserved motifs revealed that intron size, intron number, and motif number differed significantly among subfamilies. Collinearity analysis indicated that segmental duplication events were the primary driver of AhNHX family expansion. The dN/dS ratio was less than one, implying that the AhNHX gene family experienced strong purifying selection during long‑term evolution. Expression profiling across 22 tissues showed that 27 AhNHX genes exhibited tissue‑specific expression patterns. Under salt stress, transcript levels of multiple AhNHX genes were differentially regulated between the salt‑tolerant variety (HY33) and the salt‑sensitive variety (HY9115). Based on representative features in gene structure, protein sequence, evolutionary lineage, and salt‑induced upregulation, 12 genes (AhNHX11, 12, 19, 23, 34, 38, 43, 44, 48, 54, 57, and 73) were identified as candidates for further functional characterization. This study provides comprehensive insights into the AhNHX gene family and pinpoints candidate genes for improving salt tolerance in peanut via molecular breeding.

With the rapid development of new energy vehicles, the global demand for aluminum anode foils (AAF) increases continuously. In order to improve the stability and accuracy of properties prediction for AAF, a machine learning-based property prediction strategy, integrating feature selection and a stacking ensemble model, is proposed in this study. The Pearson correlation coefficient (PCC) and multicollinearity test (MT) are utilized to pinpoint the most relevant and representative features from the original datasets with 80 process parameters. The data are processed using Box-Cox transformation (BCT) to improve the normality. The stacking integrated learning model is constructed, and the Optuna algorithm is used to determine the hyperparameters of each model with the purpose of enhancing the prediction performance. The results demonstrate that the proposed model exhibits superior accuracy and stability in predicting the AAF properties in comparison to 7 baseline machine learning models. X61 (repair 2 anodizing voltage) is a key parameter of process regulation that has the strongest statistical correlation with AAF properties variation. The stepwise voltage strategy in the formation process can balance the trade-off between withstand voltage and specific capacitance. The proposed model can help to predict the two core properties of aluminum anode foils, thereby revealing the intrinsic correlations among process parameters, microstructures, and properties.

In recent years, the increasing complexity of datasets has highlighted the limitations of traditional feature selection methods and motivated the search for more stable and robust alternatives. This study proposes a bootstrap-stabilized feature selection method based on the k-means clustering algorithm, where features are repeatedly grouped, and stable representative features are identified across multiple resamples. In the proposed approach, each feature is assigned to its nearest cluster center, and the feature closest to that center is selected as the representative, reducing sensitivity to random initialization and improving generalizability. The selected features are subsequently integrated into a Fuzzy Regression Function (FRF) model, forming a hybrid framework that combines interpretable feature selection with flexible fuzzy regression modeling. The method is evaluated on ten real-world datasets and compared with widely used feature selection techniques, including Correlation based Feature Selection (CFS), Recursive Feature Elimination (RFE), Genetic Algorithm (GA) based Feature Selection, LASSO, and Ridge regression. In the proposed framework, the number of feature clusters was primarily determined using a grid-search strategy, while the Silhouette Index (SI) and Davies Bouldin Index (DBI) were additionally considered as alternative cluster validity criteria for k-means based feature selection. These findings demonstrate that the proposed bootstrap-stabilized clustering approach achieved either superior or competitive performance across the ten datasets considered, highlighting its effectiveness and suitability for complex data environments.

Materials aim to integrate excellent properties, including high strength, stiffness, significant elastic deformation, specifically at low density. However, synthetic materials usually involve trade-offs among these characteristics, resulting in distinct categories, such as hard and soft carbon materials, despite sharing identical elemental composition. Here, we demonstrate a lightweight graphene metamaterial fabricated via multi-flow assembly that integrates the mechanical robustness of low-density hard carbons with the elastic deformability of soft carbons. The representative graphene metamaterial features a cuttlebone-inspired lamella-wall architecture. This architecture reasonably strengthens and stiffens the graphene metamaterial, akin to the house-of-cards carbon layer arrangement in hard carbons. The intrinsic superelasticity under huge deformation (90%) is also retained in these graphene metamaterials. Our multi-flow assembly method is facile to prepare varied metamaterials by directly manipulating the arranged texture of individual graphene sheets, paving the way for exploring the unique properties of metamaterials in the macroscopic world and their applications.

HighlightsWhat are the main findings?Feature reconstruction effectively transfers 12-lead diagnostics to single leads.The integration of feature reconstruction and cross-attention fusion increases single-lead signal discriminability.What are the implications of the main findings?Bridging this diagnostic gap makes wearable ECGs highly reliable for daily monitoring.This strategy provides an adaptable paradigm for other limited-sensor medical domains.While the standard 12-lead ECG is vital for cardiovascular diagnosis, its reliance on clinical settings hinders daily use. Wearable few-lead devices offer a practical alternative, yet this convenience comes at the cost of diagnostic capability due to reduced lead coverage. To bridge this informational gap and enhance single-lead ECG diagnostic performance, we propose a feature-reconstruction-based classification method for single-lead ECGs. It leverages a pre-trained 12-lead ECG model to extract representative features and guide the feature learning process for single-lead signals. A CNN–Transformer-based multi-scale feature extraction module is introduced for robust ECG feature extraction, followed by a transformer encoder-based reconstruction module to align single-lead features with more discriminative 12-lead representations. A cross-attention based feature fusion module subsequently integrates the reconstructed and original single-lead features to enhance classification performance. By focusing on feature reconstruction rather than signal reconstruction, our method effectively avoids the performance degradation typically caused by signal reconstruction errors and inter-lead redundancy, leading to superior classification outcomes. Evaluation on two public datasets demonstrates that our method enhances feature discriminability and improves single-lead ECG classification performance, confirming its robustness and practical potential.

Diffusion-based models have substantially propelled the progress of portrait stylization. Nevertheless, the lack of clear supervisory signals often leads to pattern drift in the target portrait. To overcome this issue, we introduce DGADiff, a training-free stylization framework based on a diffusion model. Specifically, we first leverage prior knowledge from a pre-trained latent consistency model (LCM) to efficiently sample representative features from noisy image pairs. Next, we design a Decoupled Guide Attention Mechanism (DGA), that disentangles the U-Net attention into separate self-attention and masked-attention tracks, enabling accurate transfer of fine-grained facial style patterns. Extensive experiments verify that our DGADiff achieves favorable results across multiple metrics in content-to-style and style-to-content multi-domain tasks, demonstrating the effectiveness of spatial attention decoupling for portrait stylization.

In this study, we present the development of micro-spatially offset Raman spectroscopy (micro-SORS) methods and data analysis routines for the study of pigment degradation processes in the cultural heritage field, exploiting micro-SORS ability to non-invasively investigate the inner portions of turbid materials. The purpose of the study is to demonstrate an automated reference-free method to visualize through micro-SORS mapping the distribution of degradation both on and below the surface. The need arises from the handling of large datasets provided by micro-SORS mapping, which are often troublesome to analyse manually and usually require prior knowledge of the sample composition. Unaged and artificially aged painted mock-up samples were analysed with micro-SORS mapping, and conventional map reconstruction was compared with both supervised and unsupervised learning methods. Representative features in the micro-SORS spectra, able to distinguish unaltered pigments and degradation products, were automatically selected through machine learning techniques, revealing hidden patterns and correlations. Through the important spectral features (wavenumbers) and clustering analysis, quantitative micro-SORS degradation maps were created to identify degradation patterns also below the sample surface. Unlike previous studies that only use supervised or unsupervised learning, both are combined in this study to ensure the relevance of the selected spectral features and discover correlations among spectra through clustering techniques. This approach can be valid also for other scientific fields, such as forensic or biomedical, where data visualization and pattern identification are essential.

Abstract Accurate State-of-Charge (SOC) estimation is essential for the safe and reliable operation of lithium-ion batteries (LIBs). However, conventional SOC estimation methods mainly use electrical signals, which provide limited information about internal battery changes. Ultrasonic sensing provides additional information related to internal structural changes and has attracted increasing attention for SOC estimation. To better assess the value of ultrasonic features as auxiliary inputs for SOC estimation, this study examines them in terms of both correlation with SOC and stability under varying operating conditions. Two representative features, Time of Flight (TOF) and Signal Amplitude (SA), are systematically compared under different temperatures and C-rates. Their effectiveness is then further validated by fusing ultrasonic features with voltage, current, and temperature in a multilayer perceptron (MLP) model. The results show that SA exhibits stronger correlation with SOC and better thermal stability than TOF, and that incorporating SA into the input set of voltage, current, and temperature reduces the Root Mean Square Error (RMSE) by 29.0% under cycling conditions and 49.5% under dynamic load profiles. These results indicate that ultrasonic information can effectively improve SOC prediction performance, and the improvement is greater when more stable features are used.

Dual selective gleason pattern-aware multiple instance learning with uncertainty regularization for grade group prediction in histopathology images.

ASR-GCN: Adaptive spatial information reconstruction GCN for skeleton-based action recognition.

Single-molecule surface-enhanced Raman spectroscopy (SM-SERS), with single-molecule sensitivity and fingerprint-like specificity, has become a rising tool for probing the physicochemical properties of biomolecules at a single-molecule level. Notably, SM-SERS is generally achieved by complex methodologies with the assistance of rationally designed substrates and single-molecule techniques, often accompanied by significant technical challenges. Given the persistent issues of reproducibility and reliability, this critical perspective does not aim to provide a comprehensive review but instead progresses from a selective overview of representative SM-SERS methods and features to a detailed discussion of its intrinsic and often overlooked limitations, including the long-existing signal intensity fluctuations (SIFs), sample degradation, and the emerging challenge of big data processing and analysis. It is aimed to provide a conceptual framework that advances both the fundamental understanding and practical application of SM-SERS in future research.

The rapid growth of the elderly population worldwide demands reliable activity recognition technologies to support independent living and continuous health supervision. However, conventional wearable sensor-based human activity recognition (HAR) techniques often fail to capture the complex temporal behaviour and subtle motion patterns characteristic of the elderly. To address these limitations, this study introduces a hybrid deep residual architecture—CNN-CBAM-BiGRU—that integrates convolutional neural networks (CNNs), the convolutional block attention module (CBAM), and bidirectional gated recurrent units (BiGRUs) to improve activity recognition using inertial measurement unit (IMU) data. In the proposed CNN-CBAM-BiGRU framework, CNN layers automatically derive representative features from raw sensor signals, CBAM applies adaptive channel and spatial attention to highlight informative patterns, and BiGRU captures long-range temporal relationships within activity sequences. The approach was evaluated on three benchmark datasets designed for elderly populations—HAR70+, HARTH, and SisFall—covering daily activities and fall events. The proposed model consistently outperforms existing methods across all datasets, achieving accuracies exceeding 96%, F1-scores above 93%, and a fall detection recall of 93.74%, confirming its robustness and suitability for safety-critical monitoring applications. Class-level evaluation indicates excellent recognition of static postures and consistent performance for dynamic actions. Convergence analysis further confirms efficient learning with limited overfitting across datasets. The proposed framework thus provides a robust and accurate solution for wearable-based elderly activity recognition, with strong potential for deployment in fall detection, health monitoring, and ambient assisted living systems.