Specific learning disorder (SLD) is a neurodevelopmental disorder characterized by difficulties in reading, writing, and mathematics despite normal intelligence. One of the most common comorbidity of SLD is attention-deficit hyperactivity disorder (ADHD). However, differences between patients with SLD alone and SLD with ADHD remain underinvestigated. This study aims to compare clinical and psychosocial characteristics between these two groups. This retrospective study analyzes pediatric psychiatric patient data from Research Electronic Data Capture (REDCap) at the Learning Disorder Outpatient Clinic, Ramathibodi Hospital, Thailand from 2017 to 2023. Chi-square tests, t-tests, and mediation analyses were used to examine group differences and causal relationships among significant factors. Among 251 patients, the SLD with ADHD group exhibited less severe delays in reading (p = 0.003), writing (p = 0.001), and math (p = 0.007). This group also visited the clinic earlier (median: grade 3) than the SLD alone group (median: grade 5). Mediation analysis revealed that ADHD had a direct impact on reading and math delays (coefficients: 0.47, 0.38). No significant differences were found in gender, IQ, birth risks, parental education, or family income. Early detection explained the advantage in writing skills but not fully for reading and math, suggesting that other neurocognitive factors may be involved. Therefore, early recognition and referral should be encouraged to promote better academic outcomes.

This work aims to develop a domain-adapted Bengali text summarization model by training and fine-tuning on general and domain-specific datasets with categories such as state, international, and sports. Flan-T5 and mT5 LLMs were trained on the XLSUM Bengali dataset for text summarization, and they were trained on source domains and fine-tuned on the target domains for domain adaptation. In general-domain summarization on the XLSum Bengali dataset, mT5 achieved stronger overall performance than Flan-T5, with ROUGE-1 and BERTScore values of 0.21 and 0.72, respectively. The Flan-T5-XLSUM model outperformed other LLMs, achieving a ROUGE score of 0.79, a BLEU score of 0.13 and a BERTScore of 0.86. A human evaluation involving 28 participants was conducted to assess summary fluency, adequacy, and factual consistency across domains, confirming the qualitative effectiveness of the proposed models. A custom dataset is developed with manually annotated categories for domain adaptation purposes, contributing to domain-specific Bengali data. The ablation study illustrated that pre-training and fine-tuning on domain-specific data significantly enhanced model performance, with Flan-T5 fine-tuned on sports data. Three explainability methods (LIME, SHAP, and BertViz) revealed that geographically and contextually meaningful tokens strongly influenced the domain-adaptive Flan-T5 Bengali summarization model. The adapter-based Flan-T5 architecture enables effective multi-domain fine-tuning with less than 1% of trainable parameters while preserving model performance.

Limbal Stem Cell Deficiency (LSCD) is a corneal disorder characterized by conjunctivalization of the cornea. Central corneal epithelial thickness, measured using Anterior Segment Optical Coherence Tomography (AS-OCT), is a biomarker for LSCD. Manual measurement of epithelial thickness remains the gold standard for diagnosis because existing automated commercial software is unreliable in severe cases. Standard machine learning approaches have limited applicability for automating this measurement due to the difficulty of distinguishing corneal epithelium from scar tissue, and imprecise clinical labels. The goal of this study is to develop a deep learning pipeline to automate measurement of central epithelial thickness in AS-OCT images of eyes with LSCD. To accomplish this goal, we developed AI-CE, a model based on Multi-scale U-Net (MSU-Net) trained to accurately delineate the epithelial layer. The primary dataset comprised of 134 patient eyes, including 47 healthy normal eyes (35.0%), 34 with Stage 1 (25.4%), 34 with Stage 2 (25.4%), and 19 with Stage 3 (14.2%) LSCD. Performance was evaluated by calculating the average mean absolute error between model segmentations and manually measured clinical assessments. AI-CE model performance was compared against a commercially available segmentation software to demonstrate its improvement over the current state-of-the-art. The proposed model achieved a mean absolute error of 5.10 μm, significantly improving (p < 0.01) over clinical software (12.20 μm). This result was verified in a secondary retrospective dataset which similarly showed significant improvement over commercial software. We also demonstrate that clinical epithelial thickness values provided by AI-CE can distinguish between LSCD stages similarly to manual evaluations.

Toll-like receptor 2/1 (TLR2/1) heterodimers detect triacylated lipoproteins. Although TLR2/1 complexes localize to both the plasma membrane and endosome, the functional contributions of compartment-specific signaling are not completely defined. A common human TLR1 polymorphism (rs5743618), at which one allele substitutes isoleucine for serine adjacent to the transmembrane domain, restricts TLR1 intracellularly and is associated with an increased risk of antibiotic-refractory Lyme arthritis. This mis-localization provides a model to examine how receptor positioning influences downstream signaling. Here, we investigated the rs5743618 polymorphism's effects on TLR1 localization and compartment-specific NF-κB activation. Using a CD14+ TLR2+ HEK cell reporter system that enables transient expression of both allelic variants, we confirmed that the polymorphism is excluded from the cell surface and retained intracellularly, whereas the ancestral allele localizes to both compartments. Functionally, the ancestral allele signaled from both compartments, while the derived allele signaled exclusively from endosomes. Spatially restricted signaling exhibited influence on IkBα degradation and NF-κB activation compared with the ancestral allele. This activation was dependent on MyD88 localized at ligand-containing endosomes, establishing a role for this adapter in intracellular TLR2/1 signaling. These findings provide insight into how this common polymorphism impacts inflammatory outcomes.

Leishmania CK1.2 (L-CK1.2) is a serine/threonine protein kinase essential for the survival of the protozoan parasite Leishmania, the causative agent of leishmaniasis. This study investigates the dynamic localization pattern of L-CK1.2 and the broad spectrum of its interacting partners within the parasite. Using proteomic analysis and confocal microscopy, we identified 230 L-CK1.2-associated proteins across the major parasite life stages. This analysis revealed the ubiquitous presence of L-CK1.2 in various cellular structures, including the cytoskeleton, basal body, and flagellum. Using an in vitro system, sixty-four host proteins, involved in critical host biological processes such as immune response, apoptosis, and purine biosynthesis, were identified as potential interacting partners of L-CK1.2. These processes are known to be regulated by Leishmania during infection. The study highlights the potential dual function of L-CK1.2, in the parasite (cis) and within the host cell (trans), positioning this kinase as a key player in host-pathogen interactions. This work provides a comprehensive map of L-CK1.2 interactions and suggests its potential importance in regulating intracellular Leishmania survival, providing potential targets for leishmaniasis therapy. Furthermore, given the evolutionary conservation of CK1.2 across other parasitic organisms, our findings may have broader implications for understanding and managing parasitic infections.

Seed viability and longevity underpin forest regeneration under changing environmental conditions. European beech (Fagus sylvatica L.) produces intermediate seeds that gradually lose viability during storage, largely through oxidative stress, which makes redox homeostasis central to their longevity. Here we investigated the role of thioredoxin h1 (TRX h1), a cytosolic thiol-disulfide oxidoreductase, in the embryonic axes of beech seeds. Using a His-tagged monocysteinic (Cys to Ser) trapping mutant of TRX h1 combined with data-dependent acquisition (DDA) proteomics, we identified 161 putative TRX h1 target proteins, which fall into functional categories related to genetic information processing, metabolism, environmental and cellular signaling, and protein homeostasis. Western blotting confirmed that target capture proceeded through DTT-reducible mixed-disulfide complexes, supporting a redox-based interaction. The targets include heat shock proteins (HSP70 and HSP90) and other chaperones, calcium-binding proteins (calmodulin and calreticulin), a methyl-CpG-binding protein, glycolytic and tricarboxylic acid (TCA) cycle enzymes, and antioxidant enzymes such as superoxide dismutase, glutathione peroxidase and glutaredoxin. To link these targets to seed physiology, we quantified the thiol redox status of fresh and long-term stored seeds: total thiols were approximately 2.2-fold lower in stored than in fresh axes (21.0 ± 0.5 vs 45.9 ± 1.3 nmol GSH equivalents per axis), indicating that the reduced thiol pool, the substrate of the TRX h1 system, becomes depleted and more oxidized during storage. Together, these results suggest that TRX h1-dependent redox maintenance may contribute to redox homeostasis, energy metabolism and protein stability in beech embryonic axes, and provide a basis for improving seed storage and conservation strategies.

Emerging resistance to delamanid (DLM) and pretomanid (Pa) threatens drug-resistant tuberculosis (DR-TB) treatment outcomes. The lack of rapid molecular diagnostics, due to limited knowledge of resistance-conferring mutations, along with standardization challenges for critical concentrations (CC) in phenotypic drug susceptibility testing (pDST) hinder resistance detection. Here, we explore the correlation between minimum inhibitory concentrations (MICs) and genomic variants of DLM and Pa resistance in Mycobacterium tuberculosis (Mtb). Seventeen Mtb isolates phenotypically resistant to DLM detected through routine pDST using the Mycobacterial Growth Indicator Tube (MGIT) system from 2021-2024 were selected from the National Reference Laboratory biobank in Tbilisi, Georgia. DLM/Pa MICs were determined using the Resazurin Microtiter Assay (REMA) with a DLM concentration range of 0.008-2 µg/ml, and a broader range of 0.125-32 µg/ml was utilized for Pa. To identify mutations associated with DLM/Pa resistance, isolates were subjected to whole genome sequencing (WGS). Of the seventeen Mtb isolates phenotypically resistant to DLM, eight (47%) manifested a MIC value below the DLM CC (0.06 µg/ml), with only one carrying a mutation in a gene associated with DLM/Pa resistance (fbiC). Among the remaining nine isolates with MICs above the CC, the premature stop codon mutation W88* in ddn gene predominated, present in five cases and associated with elevated DLM and Pa MICs (0.125-2 µg/ml and 2-8 µg/ml, respectively). Three isolates (19%) harbored novel, previously undescribed genetic variants, all within genes previously linked to resistance DLM/Pa (G35fs in ddn, Q317R and Q279H in fgd1, and C565W in fbiC). Our results support DLM and Pa resistance association with W88* and W678G mutations in ddn and fbiC genes respectively, currently graded as "uncertain significance" in the WHO catalogue of Mtb drug resistance-conferring mutations. Additionally, we observed previously undescribed mutations exclusively within genes linked to DLM/Pa susceptibility, prompting further investigation of their role in resistance.

Colorectal cancer (CRC) is one of the most lethal malignancies worldwide, and the precise identification of biomarkers from colonic adenoma to cancer is of great significance for preventing the development of adenocarcinoma. Given that existing methods inadequately capture the topological network relationships among genes, this study proposes a graph neural network model based on multifeature learning, named ChebTs, to investigate the correlation between key genes involved in the colorectal "adenoma-cancer" transition. The GSE41657 and GSE31905 datasets from the GEO database were stratified into normal, adenoma, and colorectal cancer groups. Feature encoding was introduced to enhance node features, followed by Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses of differentially expressed genes (DEGs). A protein-protein interaction (PPI) network was constructed using Cytoscape software, and the aggregated information was embedded into the model for training to generate a list of key genes with corresponding importance scores. An attention pooling mechanism aggregated node-level representations into graph level representations for sample classification, and a two layer fully connected network, following activation and regularization, produced predicted probabilities. Furthermore, we provided interpretable analyses at the gene level using GNNExplainer. The results were validated through virtual knockout techniques. The eight screened genes, Fibronectin 1 (FN1), Claudin 2 (CLDN2), Interleukin-33 (IL-33), Matrix Metallopeptidase 1 (MMP1), Stanniocalcin 2 (STC2), Insulin-Like Growth Factor-Binding Protein 7 (IGFBP7), NADPH Oxidase 4 (NOX4), and Secreted Frizzled Related Protein 1 (SFRP1), were all found to be associated with overall survival (OS) in CRC. In this study, eight molecules closely related to the development of colorectal adenocarcinoma were screened out, and they may be diagnostic biomarkers of colorectal cancer. These genes affect the prognosis of patients by participating in biological processes such as remodeling of extracellular mechanisms, and are of great significance for preventing the carcinogenesis of adenoma.

Electrochemical energy storage technologies, such as lithium-ion batteries, are widely used in various scenarios, but their capacity, lifespan, and state-of-health (SOH) are susceptible to numerous influencing factors. To improve the accuracy of state-of-health (SOH) prediction for lithium-ion battery cells, this paper builds up a feature-based analysis framework, and proposes an enhanced predicting model, which integrates the principles of incremental capacity analysis (ICA) and differential voltage analysis (DVA), and the ideas of kernel principal component analysis (KPCA), and Pearson correlation analysis. The former ones are used to extract a set of health related features (HFs) from the filtered incremental capacity (IC) and differential voltage (DV) curves of batteries, while the latter ones are employed to eliminate redundant features. Furthermore, this work bridges the gap between the charging data of the cells and their SOH by building up a Newton-Raphson-based optimization Transformer-LSTM (NRBO-Transformer-LSTM) model, where NRBO algorithm is introduced to find out the optimized number of attention heads, learning rates and the regularization coefficient. Experimental validation on the NASA and CALCE battery datasets shows that, with 70% of the data used as the training set, the proposed method achieves maximum mean absolute errors (MAE) of 0.0084 and 0.0071, and root mean square errors (RMSE) of 0.0091 and 0.0089, respectively, significantly outperforming multiple benchmark algorithms in both prediction accuracy and stability.

Marital status has been associated with cognitive aging; however, whether such associations vary according to cognitive assessment modality remains unclear. This study examined whether associations between marital status and incident screen-positive cognitive impairment (MCI) differ by screening instrument and population subgroups in a large longitudinal cohort of community-dwelling older adults in Taiwan. We analyzed 103,863 adults aged ≥65 years from the Taipei City Elderly Health Examination Database (2005-2016). Cognitive status was assessed using the Short Portable Mental Status Questionnaire (SPMSQ; 2005-2011) and the 8-item Ascertain Dementia questionnaire (AD8; 2012-2016). Baseline marital status was categorized as married, divorced, widowed, or never married. Cox proportional hazards models with time-updated covariates were fitted separately for each assessment period, adjusting for demographic, socioeconomic, health, and lifestyle factors. Sensitivity analyses were conducted to examine potential reverse causation and variation in follow-up duration. Among 61,304 participants assessed during the SPMSQ period, never-married individuals had a higher risk of incident MCI compared with married individuals (hazard ratio [HR] 1.37, 95% confidence interval [CI] 1.08-1.74). In contrast, among 42,559 participants assessed during the AD8 period, never-married status was associated with a lower risk of MCI (HR 0.60, 95% CI 0.46-0.77). Effect modification by sex and age was observed across periods. Sensitivity analyses yielded consistent findings. Associations between marital status and incident MCI differed across assessment periods that varied in screening instrument, follow-up duration, and population composition; this measurement heterogeneity, rather than a confirmed temporal change, may account for the reversal observed among never-married participants. These findings suggest that interpretation of marital status-cognition associations should be considered alongside the measurement approach and study period under which they were derived.

This study presents EffAtt-LungNet, an attention-guided deep learning framework for three-class lung histopathology classification. The model combines an EfficientNetB3 backbone with the Convolutional Block Attention Module (CBAM) to refine features and improve discrimination among benign lung tissue, lung adenocarcinoma, and lung squamous cell carcinoma. Experiments used the publicly released LC25000 lung subset, which contains 15,000 images across three classes and was itself produced by augmenting a smaller collection of underlying histopathology images. We generated 10,500 additional fixed variants, creating a balanced working corpus of 25,500 images with training, validation, and test partitions of 15,300, 2,550, and 7,650 images. EffAtt-LungNet reached 99.86% accuracy, 99.86% precision, 99.86% recall, 99.86% F1 score, and 0.9997 AUC on the image-level test partition. Performance was stable across five random seeds (mean accuracy 99.86±0.02%), and paired errors differed from those of the EfficientNetB3 backbone (McNemar's test, χ2=47.27, p < 0.001). Because pre-augmentation source identifiers for the released LC25000 images were unavailable, independence at the earliest source-image level could not be confirmed. The results therefore represent internal tile-level benchmark performance and require validation on independent, non-augmented, multi-centre datasets before any translational or clinical claim can be made.