Search Results (98)

We present a semi-automatic framework for transcribing foreign personal names into Lithuanian, aimed at reducing pronunciation errors in text-to-speech systems. Focusing on noisy, web-crawled data, the pipeline combines rule-based filtering, morphological normalization, and manual stress annotation—the only non-automated step—to generate training data for character-level transcription models. We evaluate three approaches: a weighted finite-state transducer (WFST), an LSTM-based sequence-to-sequence model with attention, and a Transformer model optimized for character transduction. Results show that word-pair models outperform single-word models, with the Transformer achieving the best performance (19.04% WER) on a cleaned and augmented dataset. Data augmentation via word order reversal proved effective, while combining single-word and word-pair training offered limited gains. Despite filtering, residual noise persists, with 54% of outputs showing some error, though only 11% were perceptually significant. Full article

Chemiresistive gas sensors have gained significant attention and have been widely applied in various fields. However, the gap between experimental observations and fundamental sensing mechanisms hinders systematic optimization. Despite the critical role of modeling in explaining atomic-scale interactions and offering predictive insights beyond experiments, existing reviews on chemiresistive gas sensors remain predominantly experimental-centric, with a limited systematic exploration of the modeling approaches. Herein, we present a comprehensive overview of the modeling approaches for chemiresistive gas sensors, focusing on two critical processes: the reception and transduction stages. For the reception process, density functional theory (DFT), molecular dynamics (MD), ab initio molecular dynamics (AIMD), and Monte Carlo (MC) methods were analyzed. DFT quantifies atomic-scale charge transfer, and orbital hybridization, MD/AIMD captures dynamic adsorption kinetics, and MC simulates equilibrium/non-equilibrium behaviors based on statistical mechanics principles. For the transduction process, band-bending-based theoretical models and power-law models elucidate the resistance modulation mechanisms, although their generalizability remains limited. Notably, the finite element method (FEM) has emerged as a powerful tool for full-process modeling by integrating gas diffusion, adsorption, and electronic responses into a unified framework. Future directions highlight the use of multiscale models to bridge microscopic interactions with macroscopic behaviors and the integration of machine learning, accelerating the iterative design of next-generation sensors with superior performance. Full article

Background: Membranous nephropathy (MN), a prevalent glomerular disorder, remains poorly understood in terms of its association with mitochondrial dynamics (MD). This study investigated the mechanistic involvement of mitochondrial dynamics-related genes (MDGs) in the pathogenesis of MN. Methods: Comprehensive bioinformatics analyses—encompassing Mendelian randomization, machine-learning algorithms, and single-cell RNA sequencing (scRNA-seq)—were employed to interrogate transcriptomic datasets (GSE200828, GSE73953, and GSE241302). Core MDGs were further validated using reverse-transcription quantitative polymerase chain reaction (RT-qPCR). Results: Four key MDGs—RTTN, MYO9A, USP40, and NFKBIZ—emerged as critical determinants, predominantly enriched in olfactory transduction pathways. A nomogram model exhibited exceptional diagnostic performance (area under the curve [AUC] = 1). Seventeen immune cell subsets, including regulatory T cells and activated dendritic cells, demonstrated significant differential infiltration in MN. Regulatory network analyses revealed ATF2 co-regulation mediated by RTTN and MYO9A, along with RTTN-driven modulation of ELOA-AS1 via hsa-mir-431-5p. scRNA-seq analysis identified mesenchymal–epithelial transitioning cells as key contributors, with pseudotime trajectory mapping indicating distinct temporal expression profiles: NFKBIZ (initial upregulation followed by decline), USP40 (gradual fluctuation), and RTTN (persistently low expression). RT-qPCR results corroborated a significant downregulation of all four genes in MN samples compared to controls (p < 0.05). Conclusions: These findings elucidate the molecular underpinnings of MDG-mediated mechanisms in MN, revealing novel diagnostic biomarkers and therapeutic targets. The data underscore the interplay between mitochondrial dynamics and immune dysregulation in MN progression, providing a foundation for precision medicine strategies. Full article

Malware continues to evolve rapidly, posing significant challenges to network security. Traditional signature-based detection methods often struggle to cope with advanced evasion techniques such as polymorphism, metamorphism, encryption, and stealth, which are commonly employed by cybercriminals. As a result, these conventional approaches frequently fail to detect newly emerging malware variants in a timely manner. To address this limitation, Zero-Shot Learning (ZSL) has emerged as a promising alternative, offering improved classification capabilities for previously unseen malware samples. ZSL models leverage auxiliary semantic information and binary feature representations to enhance the recognition of novel threats. This study proposes a Transductive Zero-Shot Learning (TZSL) model based on the Vector Quantized Variational Autoencoder (VQ-VAE) architecture, integrated with a malware knowledge graph constructed from sandbox behavioral analysis of ransomware families. The model is further optimized through hyperparameter tuning to maximize classification performance. Evaluation metrics include per-family classification accuracy, precision, recall, F1-score, and Receiver Operating Characteristic (ROC) curves to ensure robust and reliable detection outcomes. In particular, the harmonic mean (H-mean) metric from the Generalized Zero-Shot Learning (GZSL) framework is introduced to jointly evaluate the model’s performance on both seen and unseen classes, offering a more holistic view of its generalization ability. The experimental results demonstrate that the proposed VQ-VAE model achieves an F1-score of 93.5% in ransomware classification, significantly outperforming other baseline models such as LeNet-5 (65.6%), ResNet-50 (71.8%), VGG-16 (74.3%), and AlexNet (65.3%). These findings highlight the superior capability of the VQ-VAE-based TZSL approach in detecting novel malware variants, improving detection accuracy while reducing false positives. Full article

As the global elderly population is rising, concerns about cognitive decline and memory loss are becoming urgent. This study evaluated the potential of sea cucumber hydrolysates (SCH) from Stichopus japonicus in alleviating cognitive deficits using a D-galactose-induced murine aging model. The effects of SCH on behavior, hippocampal morphology, gut microbiota, hippocampal cholinergic system, brain-derived neurotrophic factor (BDNF) signaling, and neuroinflammatory pathways were investigated. Results showed that SCH ameliorated learning and memory deficits and reduced neuronal damage in aging mice. SCH also modulated gut microbiota, along with increased fecal short-chain fatty acids levels. Functional prediction revealed that alterations in gut microbiota were related to signal transduction. Further, SCH enhanced hippocampal cholinergic function through elevating acetylcholine (ACh) levels and inhibiting acetylcholinesterase (AChE) activity and activated BDNF signaling, consistent with predictions of gut microbiota function. Restoration of cholinergic homeostasis and transmission of the BDNF pathway might contribute to the inhibition of hippocampal neuroinflammation via suppressing microglial activation and the nuclear factor kappa-B (NF-κB) pathway. In summary, SCH attenuated cognitive deficits through suppressing neuroinflammation, which might be correlated with the signal transduction caused by regulating gut microbiota. Further validation will be conducted through microbiota depletion and fecal microbiota transplantation. These findings suggest that SCH is a promising functional component for counteracting aging-related cognitive deficits. Full article

Doping undermines fairness in sports and threatens athlete health, while conventional detection methods like LC-MS and GC-MS face challenges such as complex procedures, matrix interferences, and lengthy processing times, limiting on-site applications. Two-dimensional (2D) materials, including graphene, MoS₂, and metal–organic frameworks (MOFs), offer promising solutions due to their large surface areas, tunable electronic structures, and special interactions with doping agents, such as hydrogen bonding, π-π stacking, and electrostatic forces. These materials enable signal transduction through changes in conductivity or fluorescence quenching. This review highlights the use of 2D materials in doping detection. For example, reduced graphene oxide–MOF composites show high sensitivity for detecting anabolic steroids like testosterone, while NiO/NGO nanocomposites exhibit strong selectivity for stimulants like ephedrine. However, challenges such as environmental instability and high production costs hinder their widespread application. Future efforts should focus on improving material stability through chemical modifications, reducing production costs, and integrating these materials into advanced systems like machine learning. Such advancements could revolutionize doping detection, ensuring fairness in sports and protecting athlete health. Full article

Background/Objectives: Genomic prediction is a powerful approach that predicts phenotypic traits from genotypic information, enabling the acceleration of trait improvement in plant breeding. Traditional genomic prediction methods have primarily relied on linear mixed models, such as Genomic Best Linear Unbiased Prediction (GBLUP), and conventional machine learning methods like Support Vector Regression (SVR). Traditional methods are limited in handling high-dimensional data and nonlinear relationships. Thus, deep learning methods have also been applied to genomic prediction in recent years. Methods: We proposed iADEP, Integrated Additive, Dominant, and Epistatic Prediction model based on deep learning. Specifically, single nucleotide polymorphism (SNP) data integrating latent genetic interactions and genome-wide association study results as biological prior knowledge are fused to an SNP embedding block, which is then input to a local encoder. The local encoder is fused with an omic-data-incorporated global decoder through a multi-head attention mechanism, followed by multilayer perceptrons. Results: Firstly, we demonstrated through experiments on four datasets that iADEP outperforms existing methods in genotype-to-phenotype prediction. Secondly, we validated the effectiveness of SNP embedding through ablation experiments. Third, we provided an available module for combining other omics data in iADEP and propose a novel method for fusing them. Fourthly, we explored the impact of feature selection on iADEP performance and conclude that utilizing the full set of SNPs generally provides optimal results. Finally, by altering the partition of training and testing sets, we investigated the differences between transductive learning and inductive learning. Conclusions: iADEP provides a new approach for AI breeding, a promising method that integrates biological prior knowledge and enables combination with other omics data. Full article

Proteins are essential for all living organisms, playing key roles in biochemical reactions, structural support, signal transduction, and gene regulation. Their importance in biomedical research is highlighted by their role as drug targets in various diseases. The interactions between proteins and nanoparticles (NPs), including the protein corona’s formation, significantly affect NP behavior, biodistribution, cellular uptake, and toxicity. Comprehending these interactions is pivotal for advancing the design of NPs to augment their efficacy and safety in biomedical applications. While traditional nanomedicine design relies heavily on experimental work, the use of data science and machine learning (ML) is on the rise to predict the synthesis and behavior of nanomaterials (NMs). Nanoinformatics combines computational simulations with laboratory studies, assessing risks and revealing complex nanobio interactions. Recent advancements in artificial intelligence (AI) and ML are enhancing the characterization of the protein corona and improving drug discovery. This review discusses the advantages and limitations of these approaches and stresses the importance of comprehensive datasets for better model accuracy. Future developments may include advanced deep-learning models and multimodal data integration to enhance protein function prediction. Overall, systematic research and advanced computational tools are vital for improving therapeutic outcomes and ensuring the safe use of NMs in medicine. Full article

Transient receptor potential vanilloid 4 (TRPV4) is a calcium-permeable cation channel critical for maintaining intracellular Ca²⁺ homeostasis and is essential in regulating immune responses, metabolic processes, and signal transduction. Recent studies have shown that TRPV4 activation enhances influenza A virus infection, promoting viral replication and transmission. However, there has been limited exploration of antiviral drugs targeting the TRPV4 channel. In this study, we developed the first machine learning model specifically designed to predict TRPV4 inhibitory small molecules, providing a novel approach for rapidly identifying repurposed drugs with potential antiviral effects. Our approach integrated machine learning, virtual screening, data analysis, and experimental validation to efficiently screen and evaluate candidate molecules. For high-throughput virtual screening, we employed computational methods to screen open-source molecular databases targeting the TRPV4 receptor protein. The virtual screening results were ranked based on predicted scores from our optimized model and binding energy, allowing us to prioritize potential inhibitors. Fifteen small-molecule drugs were selected for further in vitro and in vivo antiviral testing against influenza. Notably, glecaprevir and everolimus demonstrated significant inhibitory effects on the influenza virus, markedly improving survival rates in influenza-infected mice (protection rates of 80% and 100%, respectively). We also validated the mechanisms by which these drugs interact with the TRPV4 channel. In summary, our study presents the first predictive model for identifying TRPV4 inhibitors, underscoring TRPV4 inhibition as a promising strategy for antiviral drug development against influenza. This pioneering approach lays the groundwork for future clinical research targeting the TRPV4 channel in antiviral therapies. Full article

Background: Gynecomastia, caused by an estrogen–testosterone imbalance, affects males across various age groups. With unclear mechanisms and no approved drugs, the condition underscores the need for efficient, innovative treatment strategies. Methods: This study utilized deep learning-based computational methods to discover potential drug compounds for gynecomastia. To identify genes and pathways associated with gynecomastia, initial analyses included text mining, biological process exploration, pathway enrichment and protein–protein interaction (PPI) network construction. Subsequently, drug–target interactions (DTIs) were examined to identify potential therapeutic compounds. The DeepPurpose toolkit was employed to predict interactions between these candidate drugs and gene targets, prioritizing compounds based on their predicted binding affinities. Results: Text mining identified 177 genes associated with gynecomastia. Gene Ontology (GO) biological process and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analyses identified critical genes and pathways, with notable involvement in signal transduction, cell proliferation and steroid hormone biosynthesis. PPI network analysis highlighted 10 crucial genes, such as IGF1, TGFB1 and AR. DTI analysis and DeepPurpose predictions identified 12 potential drugs, including conteltinib, yifenidone and vosilasarm, with high predicted binding affinities to the target genes. Conclusions: The study successfully identified potential drug compounds for gynecomastia using a deep learning-based approach. The findings highlight the effectiveness of combining text mining and artificial intelligence in drug discovery. This innovative method provides a new avenue for developing specific treatments for gynecomastia and underscores the need for further experimental validation and optimization of prediction models to support novel drug development. Full article

Coronary artery disease (CAD), the build-up of atherosclerotic plaques on the wall of blood vessels, causes adverse cardiovascular events. Secondary prevention focuses on treating patients with existing plaques to prevent disease progression. Recent studies have shown that inflammation is an independent risk factor that drives disease progression, and targeting inflammation could be an effective therapeutic strategy for secondary prevention. In this review, we highlighted the roles of several inflammatory pathways in rupture and erosion, two major processes through which established plaques lead to adverse cardiovascular events. In the past 15 years, numerous clinical trials have tested the therapeutic potential of targeting these pathways, including neutralizing inflammatory cytokines and blocking signaling transduction of the inflammatory pathways. Only colchicine was approved for clinical use in patients with CAD. This is primarily due to the multifaceted roles of inflammatory pathways in disease progression. Commonly used pre-clinical models provided robust information for the onset of early disease but limited understanding of the inflammatory network in established plaques. This review will summarize lessons learned from successful and failed clinical trials to advocate for assessing inflammation in established plaques before designing therapeutics for secondary prevention. Full article

The src-homology 2 domain-containing phosphatase 2 (SHP2) is a human cytoplasmic protein tyrosine phosphatase that plays a crucial role in cellular signal transduction. Aberrant activation and mutations of SHP2 are associated with tumor growth and immune suppression, thus making it a potential target for cancer therapy. Initially, researchers sought to develop inhibitors targeting SHP2’s catalytic site (protein tyrosine phosphatase domain, PTP). Due to limitations such as conservativeness and poor membrane permeability, SHP2 was once considered a challenging drug target. Nevertheless, with the in-depth investigations into the conformational switch mechanism from SHP2’s inactive to active state and the emergence of various SHP2 allosteric inhibitors, new hope has been brought to this target. In this study, we investigated the interaction models of various allosteric inhibitors with SHP2 using molecular dynamics simulations. Meanwhile, we explored the free energy landscape of SHP2 activation using enhanced sampling technique (meta-dynamics simulations), which provides insights into its conformational changes and activation mechanism. Furthermore, to biophysically interpret high-dimensional simulation trajectories, we employed interpretable machine learning methods, specifically extreme gradient boosting (XGBoost) with Shapley additive explanations (SHAP), to comprehensively analyze the simulation data. This approach allowed us to identify and highlight key structural features driving SHP2 conformational dynamics and regulating the activity of the allosteric inhibitor. These studies not only enhance our understanding of SHP2’s conformational switch mechanism but also offer crucial insights for designing potent allosteric SHP2 inhibitors and addressing drug resistance issues. Full article

The Transductive Support Vector Machine (TSVM) is an effective semi-supervised learning algorithm vulnerable to adversarial sample attacks. This paper proposes a new adversarial attack method called the Multi-Stage Dual-Perturbation Attack (MSDPA), specifically targeted at TSVMs. The MSDPA has two phases: initial adversarial samples are generated by an arbitrary range attack, and finer attacks are performed on critical features to induce the TSVM to generate false predictions. To improve the TSVM’s defense against MSDPAs, we incorporate adversarial training into the TSVM’s loss function to minimize the loss of both standard and adversarial samples during the training process. The improved TSVM loss function considers the adversarial samples’ effect and enhances the model’s adversarial robustness. Experimental results on several standard datasets show that our proposed adversarial defense-enhanced TSVM (adv-TSVM) performs better in classification accuracy and adversarial robustness than the native TSVM and other semi-supervised baseline algorithms, such as S³VM. This study provides a new solution to improve the defense capability of kernel methods in an adversarial setting. Full article

Calcium ions play a key role in the physiological processes of the central nervous system. The intracellular calcium signal, in nerve cells, is part of the neurotransmission mechanism. They are responsible for stabilizing membrane potential and controlling the excitability of neurons. Calcium ions are a universal second messenger that participates in depolarizing signal transduction and contributes to synaptic activity. These ions take an active part in the mechanisms related to memory and learning. As a result of depolarization of the plasma membrane or stimulation of receptors, there is an extracellular influx of calcium ions into the cytosol or mobilization of these cations inside the cell, which increases the concentration of these ions in neurons. The influx of calcium ions into neurons occurs via plasma membrane receptors and voltage-dependent ion channels. Calcium channels play a key role in the functioning of the nervous system, regulating, among others, neuronal depolarization and neurotransmitter release. Channelopathies are groups of diseases resulting from mutations in genes encoding ion channel subunits, observed including the pathophysiology of neurological diseases such as migraine. A disturbed ability of neurons to maintain an appropriate level of calcium ions is also observed in such neurodegenerative processes as Alzheimer’s disease, Parkinson’s disease, Huntington’s disease, and epilepsy. This review focuses on the involvement of calcium ions in physiological and pathological processes of the central nervous system. We also consider the use of calcium ions as a target for pharmacotherapy in the future. Full article

Few-shot learning (FSL) is a learning paradigm which aims to address the issue of machine/deep learning techniques which traditionally need huge amounts of labelled data to work out. The remote sensing (RS) community has explored this paradigm with numerous published studies to date. Nevertheless, there is still a need for clear pieces of evidence on FSL-related issues in the RS context, such as which of the inference approaches is more suitable: inductive or transductive? Moreover, how does the number of epochs used during training, based on the meta-training (base) dataset, relate to the number of unseen classes during inference? This study aims to address these and other relevant questions in the context of FSL for scene classification in RS images. A comprehensive evaluation was conducted considering eight FSL approaches (three inductive and five transductive) and six scene classification databases. Some conclusions of this research are as follows: (1) transductive approaches are better than inductive ones. In particular, the transductive technique Transductive Information Maximisation (TIM) presented the best overall performance, where in 20 cases it got the first place; (2) a larger number of training epochs is more beneficial when there are more unseen classes during the inference phase. The most impressive gains occurred particularly considering the AID (6-way) and RESISC-45 (9-way) datasets. Notably, in the AID dataset, a remarkable 58.412% improvement was achieved in 1-shot tasks going from 10 to 200 epochs; (3) using five samples in the support set is statistically significantly better than using only one; and (4) a higher similarity between unseen classes (during inference) and some of the training classes does not lead to an improved performance. These findings can guide RS researchers and practitioners in selecting optimal solutions/strategies for developing their applications demanding few labelled samples. Full article