Personalized therapy in acromegaly is limited by interindividual variability in drug responses and the lack of robust markers predicting tumor shrinkage, rather than biochemical control alone. To test whether ex vivo drug-induced viability changes in patient-derived 3D cultures (Pd3D) of GH-secreting pituitary adenomas reflect tumor cell-intrinsic pharmacological sensitivity and align with established clinical predictors. Spheroid-based Pd3D cultures were established from 27 patients with acromegaly. Cultures were exposed to octreotide, cabergoline, pasireotide, or vehicle control. We assessed cell viability changes; sample-level responder status (viability reduction vs vehicle, p < 0.05); and associations between responder status and known predictive markers, including clinical characteristics, MRI findings, dynamic drug tests, and pathological features. In 6 cases, AI-based digital image analysis quantified pre- and post-treatment SSTR2 expression in liquid-based cytology (LBC). All agents modestly reduced median cell viability (84-86%, p < 0.05), with responder rates of 33-40%. Concordance with established predictors was observed: octreotide responders correlated with T2 hypointensity (88% vs 44%, p = 0.04); cabergoline with positive bromocriptine tests (100% vs 45%, p = 0.03); and pasireotide with sparsely granulated patterns (64% vs 19%, p = 0.04). AI-based dynamic analysis demonstrated that ex vivo responders showed relatively stable SSTR2 expression after treatment, whereas nonresponders exhibited marked depletion. Spheroid-based Pd3D ex vivo viability assays revealed modest but significant cohort-level effects and sample-level concordance with clinical predictors, suggesting its potential utility as an exploratory model. Additionally, AI-based quantification of SSTR2 dynamics captured functional receptor shifts.
Cancer progresses through the accumulation of genomic events. Cancer progression models such as Mutual Hazard Networks (MHNs) describe this dynamic, enabling prediction of temporal event positions and patient-specific risks of acquiring mutations. However, current MHN analyses rely on single most likely models and do not quantify the uncertainty inherent to parameter estimation. Assessing forecast stability is essential before using them to anticipate treatment-relevant mutations, adapt targeted therapies, or prioritize monitoring of patients at elevated progression risk. We address a key prerequisite for the responsible clinical use of cancer progression models by making MHN-derived predictions uncertainty-aware. We present a Bayesian framework for MHN that uses Markov Chain Monte Carlo to sample from the posterior distributions of model parameters and derived predictions. For practical use we implemented the Random-Walk Metropolis, Metropolis-Adjusted Langevin Algorithm (MALA) and simplified manifold MALA samplers as part of the existing mhn Python package. Only MALA and smMALA were successful in sampling from MHN posteriors, with MALA performing best. While most MHN parameters and predictions showed low posterior variance, a small subset displayed greater variability across the posterior distribution. This differentiation cannot be obtained from a single most likely model, emphasizing the need for uncertainty quantification, especially in clinical contexts. As an illustrative example, posterior sampling identified a subgroup of STK11$-$, KRAS$+$ lung adenocarcinoma patients with a high predicted short-term risk - with low variance across posterior samples - to develop an STK11 mutation. This subgroup exhibited poorer survival under immunotherapy, resembling patterns observed in STK11+ patients. Our implementation is part of version 1.2.0 of the mhn package (https://github.com/spang-lab/LearnMHN). All analyses including the code to produce all figures in this article can be found under https://github.com/huy29433/MCMC-sampling-for-MHN (https://doi.org/10.5281/zenodo.21160219).
This research aims to describe the flow characteristics and entropy creation of an Oldroyd-B tri hybrid nanofluid under conditions of MHD with heat transfer through a hybrid numerical-machine learning framework. The nonlinear boundary layer governing equations for momentum and heat transport are transformed to a commonly used ordinary differential equations (ODE) form via similarity transformations and solved using MATLAB's bvp4c solver. A mesh independence study has verified numerical. The main innovation of this study is combining ANN modelling with the nonlinear numerical calculation of Oldroyd-B tri hybrid nanofluid flow to create an accurate predictive surrogate modelling tool for complex thermofluid systems. Data collected from the bvp4c solver was then used to train a feed forward ANN, using Levenberg-Marquardt backpropagation algorithm. The trained ANN was able to accurately predict velocity, temperature and entropy creation, with regression accuracies above 0.999 and mean square error values less than [Formula: see text], indicating a high degree of predictive power. Parametric analysis indicated that increasing the magnetic parameter caused a large reduction in velocity field (approximately 18-25%) as a result of the Lorentz force acting in against the flow. Also, the generation of heat increased the temperature profile by approximately 20%, therefore increasing Entropy Generation within the thermal boundary layer. In addition, tri-hybrid nanoparticles have better thermal conductivity and heat transfer performance than nanofluids made from conventional materials because of their ability to improve the thermal performance of fluids. A predictive framework for ANN-based modelling of nonlinear fluid transport has been developed that reduces the computational cost of obtaining accurate numerical solutions compared with traditional methods. This new framework has the potential to allow engineers and scientists to model advanced nanofluids with greater accuracy than previous approaches, thereby providing valuable information for thermal management, high-performance cooling systems, and energy conversion technologies.
Additional evaluation after pediatric febrile urinary tract infection (UTI) is required to identify children at risk of long-term renal complications. Because these complications are difficult to directly assess in clinical practice, we sought to identify clinical predictors that may guide additional evaluation, using subsequent intervention as a pragmatic indicator of clinically significant urological concern following a first febrile UTI. We retrospectively reviewed 2,155 children aged < 10 years with a first febrile UTI. Clinical factors at diagnosis were analyzed to identify predictors associated with subsequent intervention. The sensitivity and specificity of renal and bladder ultrasonography (RBUS) for predicting subsequent intervention were evaluated when performed universally or selectively according to these predictors. Of the 2,155 children, 88 (4.1%) subsequently underwent intervention, including continuous antibiotic prophylaxis in 40 and surgical procedures in 48. Hospitalization or antibiotic use within the preceding month (p < 0.001), age ≥ 24 months (p < 0.001), infection with a non-E.coli pathogen (p < 0.001), failure to achieve defervescence ≤ 2 days after antibiotic therapy (p = 0.038), and peak body temperature > 39.0 ℃ (p = 0.007) were independently associated with subsequent intervention. Among the 2,047 children who underwent RBUS, 993 (48.5%) had at least one identified risk factor. Universal (n = 2,047) and selective (n = 993) RBUS demonstrated similar sensitivity and specificity for identifying children requiring intervention, while selective use reduced RBUS utilization by 51.5%. Clinical factors identified at diagnosis may help guide selective additional evaluation in children after their first febrile UTI.
Piperacillin/tazobactam (TZP) is widely used in hospitalised adults. Real‑world data suggest a higher risk of hypokalaemia than previously recognised. To investigate incidence, severity and risk factors of TZP‑induced hypokalaemia, and to develop and externally validate a nomogram for individualised risk prediction. This retrospective study included hospitalised adults receiving TZP (2021-2025). TZP‑induced hypokalaemia was defined as serum potassium < 3.5 mmol/L ≥ 48 h after TZP initiation. Logistic regression identified predictors. A nomogram was constructed. Model performance was assessed by discrimination (area under the ROC curve, AUC), calibration (Hosmer‑Lemeshow test, calibration intercept/slope), Brier score and decision‑curve analysis (DCA), with internal and external validation. Among 643 patients, 122 (19.0%) developed TZP‑induced hypokalaemia (mild: 78.7%; moderate: 16.4%; severe: 4.9%). Median time to onset was 5.0 days. Independent predictors were age ≥ 65 years (OR = 2.118, 95% CI 1.003-4.472, p = 0.049), body mass index (OR = 0.671, 95% CI 0.591-0.762, p < 0.001), intensive care unit admission (OR = 2.915, 95% CI 1.504-5.651, p = 0.002), daily dose (OR = 1.280, 95% CI 1.110-1.476, p < 0.001), and baseline serum potassium (OR = 0.123, 95% CI 0.056-0.271, p < 0.001). The nomogram showed good discrimination (training AUC 0.881, internal 0.876, external 0.813), calibration (Hosmer‑Lemeshow p > 0.05), Brier scores < 0.25 and positive net benefit on DCA. TZP‑induced hypokalaemia is common in hospitalised adults. A nomogram based on five readily available clinical variables may facilitate early identification of high‑risk patients, although prospective validation in diverse settings is required before clinical implementation.
Robotic-assisted surgery has transformed minimally invasive surgery worldwide, offering improved precision, visualization, and dexterity. However, evidence describing program implementation and perioperative outcomes in low- and middle-income countries remains limited. This study describes the initial institutional experience of a multidisciplinary robotic-assisted surgery program in Peru. A retrospective cohort study included consecutive adult patients undergoing robotic-assisted surgery during the early implementation phase (December 1, 2024 to March 31, 2026) at a national referral center in Peru. Demographic, clinical, procedural, and perioperative outcomes were collected. Multivariable linear regression identified independent predictors of operative time. A total of 398 robotic-assisted procedures were analyzed. Mean age was 49.8 ± 17.2 years, and 68.2% of patients were female. Most procedures were performed for benign disease (64.8%), followed by malignancy (29.4%) and complex/reconstructive indications (3.8%). Cholecystectomy (18.1%) and hysterectomy (17.6%) were the most common procedures. Mean operative time was 236.8 ± 152.1 min. The conversion rate was 7.5%, postoperative complications occurred in 10.3% of patients, mean length of stay was 7.68 ± 8.17 days, 30-day readmission was 2.5%, and no postoperative mortality occurred. Increasing age (β = 0.85 min/year; p = 0.032), male sex (β = 52.3 min; p = 0.011), and higher ASA classification (β = 8.7 min/category; p < 0.001) independently predicted longer operative time. The model explained 54% of operative-time variability (R²=0.54). Early implementation of a multidisciplinary robotic-assisted surgery program in a Peruvian public referral hospital demonstrated broad procedural adoption, acceptable perioperative outcomes, and predictable variation in operative duration. These findings provide an implementation benchmark for robotic-assisted surgery in resource-limited healthcare systems.
Personalized tools for accurately predicting papillary thyroid carcinoma (PTC) patients' response to initial 131I therapy are lacking. This study aimed to develop a machine learning (ML) model for prediction of 6-12 month therapeutic response after therapy. This study retrospectively enrolled 702 PTC patients undergoing initial 131I therapy. A hierarchical classification framework was designed: the first layer distinguished excellent response from non-excellent response (Non-ER), while the second layer further categorized Non-ER into indeterminate response, biochemical incomplete response, and structural incomplete response. Seven ML algorithms were utilized per layer. Feature selection was performed using recursive feature elimination. Model development employed nested five-fold cross-validation within the training set, with performance further evaluated on an independent testing set. The final model was benchmarked against the American Thyroid Association risk stratification system, interpreted via Shapley Additive exPlanations, and deployed as an interactive web tool. For the first-layer classification, the Random Forest model optimally identified Non-ER. For the second-layer classification, the Logistic Regression + Extreme Gradient Boosting fusion model excelled. The final model, built on 17 features, achieved a hierarchical F1 score of 0.849 and an overall accuracy of 69.7%. Both the model's superior predictive accuracy over the benchmark model and its strong association with progression-free survival confirmed the value of this hierarchical model. This study developed an ML-based hierarchical model capable of forecasting early response to initial 131I therapy in PTC, offering a practical web-based tool for personalized clinical decision-making.
Despite the rise of cancer immunotherapy, cisplatin remains a cornerstone in chemotherapy. Hyponatremia is a common but under-recognized adverse event of cisplatin. Previous studies on cisplatin-induced hyponatremia (CIH) are limited and mostly focus on the initial administration. This study aimed to assess the overall incidence of CIH and identify significant predictive factors. We conducted a single-center, retrospective cohort study of patients who received intravenous cisplatin at Juntendo University Hospital between 2018 and 2020. The occurrence and severity of CIH were assessed across all administrations. Predictive factors were analyzed using multiple logistic regression and generalized linear mixed models. Among 706 patients, 340 (48.2%) experienced CIH during the study period, with 103 (14.6%) developing Grade 3 or higher severity. Across 1939 total cisplatin administrations, CIH occurred in 585 (30.2%) cases, including 135 (7.0%) with Grade ≥ 3 severity. CIH was most common at first administration (38.8%), with severe cases occurring in 11.9%. Generalized linear mixed models results that the significant prognostic factors for cisplatin-induced Hyponatremia were age (OR: 1.41, p < 0.001), last cisplatin dose (OR: 1.74, p < 0.001), serum albumin (OR: 0.65, p < 0.001), serum sodium (OR: 0.35, p < 0.001), WBC (OR: 1.26, p = 0.01), and concomitant 5-FU administration (OR: 3.91, p < 0.001). CIH is a frequent and potentially serious complication of cisplatin therapy. Older age, high-dose cisplatin, hypoalbuminemia, hyponatremia, leukocytosis, and 5-FU co-administration significantly increase the risk. These findings highlight the importance of proactive monitoring and management strategies to mitigate CIH risk.
In this article, we present a computational framework for predicting treatment response to neurofeedback (NF) among patients with Attention-Deficit/Hyperactivity Disorder (ADHD). The proposed framework uses functional brain connectivity analysis of electroencephalogram (EEG) signals acquired during an early-to-mid NF treatment window to classify participants as eventual responders or non-responders. The six-stage algorithm was evaluated using an open-access EEG dataset from the Mendeley Data repository comprising 60 children with ADHD aged 6-12 years. The framework includes a preprocessing pipeline designed to reduce EEG artifacts and noise. Next, spectral features, specifically of the alpha and beta frequency bands, were extracted from the noise-reduced signals. In the fourth stage, functional connectivity was estimated by calculating Phase Locking Value (PLV) between all electrode pairs, thereby quantifying inter-channel phase synchronization. The fifth stage, which is an essential stage, was dimensionality reduction to find the most discriminative features. Dimensionality reduction was achieved in a two-process manner; for the first process, statistical screening was performed using Welch's t-test with FDR correction, followed by GA-based channel selection to identify the most discriminative electrode subset. The GA analysis identified a compact six-channel subset consisting of C3, C4, Cz, Fz, Fp1, and T6 from the original 32-channel montage. In the last classification stage, the reduced feature vectors were input as part of an ensemble of machine learning classifiers to achieve classification. Model performance was evaluated using subject-wise grouped cross-validation, with the best configuration achieving an accuracy of 84.72%. These results suggest that the proposed data-driven framework may support future research on individualized NF response prediction, pending validation on independent clinical datasets.
Postoperative infectious complications (POI) after gastrectomy remain a major concern. While bioelectrical impedance analysis (BIA) is used for nutritional assessment, the clinical relevance of the preoperative extracellular water-to-total body water ratio (ECW/TBW) remains unclear. We investigated the association between preoperative ECW/TBW and POI. This single-center retrospective cohort study included 181 patients who underwent curative gastrectomy between November 2020 and October 2025. Preoperative ECW/TBW was measured using multifrequency BIA. The primary outcome was POI, defined as surgical site infection (SSI) and/or remote infection (RI) within 30 days. Multivariable logistic regression and receiver operating characteristic (ROC) analyses were performed. POI developed in 44 patients (24.3%), as SSI in 30 (16.6%) and as RI in 16 (8.8%). Multivariable analysis identified that a higher ECW/TBW was associated with POI (odds ratio [OR] per 0.01 increase, 1.68; 95% confidence interval [CI], 1.02-2.77; P = 0.040) and RI (OR 1.79; 95% CI, 1.00-3.21; P = 0.048), but not with SSI. ROC analysis showed a numerically higher AUC of ECW/TBW for predicting RI (area under the curve, 0.833) than phase angle (0.761). Elevated ECW/TBW may reflect systemic physiological vulnerability associated with RI, particularly postoperative respiratory infection.
暂无摘要（点击查看详情）
Molecular scoring is a popular approach for studying pathway-level functional alterations with omics data. Using molecular scores for tasks such as single-sample molecular characterisation, phenotype prediction or disease stratification has several advantages compared to using omics data directly. Molecular scores provide biological interpretability and are more generalisable across datasets, facilitating data integration and machine learning applications. However, numerous scoring methods are available through different software packages, and currently there is a lack of tools to easily use these scores for model training and prediction. We developed pathMED, an R/Bioconductor package that unifies various scoring methods in a simple framework. Furthermore, pathMED also contains a machine learning module to train and test models that use the calculated molecular scores to predict clinical outcomes. We demonstrate some of its potential applications in three use cases using public omics data. We showed the generalisability of machine learning models trained on transcriptomic scores in predicting clinical outcomes when deploying on proteomic scores. We also demonstrated the application of transcriptomics scores in predicting breast cancer treatment response and identifying pathways strongly associated to tumour biology and treatment response. Finally, we demonstrated the benefit of integrating a novel gene set dissection step into the analysis pipeline to resolve disease heterogeneity at the pathway level. PathMED is freely available in the Bioconductor repository (https://bioconductor.org/packages/release/bioc/html/pathMED.html). Code to reproduce the analyses is publicly available at https://github.com/GENyO-BioInformatics/pathMED_article. Supplementary data are available at Bioinformatics online.
Accurate prognostication after aneurysmal subarachnoid hemorrhage (aSAH) remains challenging. Conventional clinical and radiological grading systems, including the World Federation of Neurosurgical Societies (WFNS), Hunt-Hess, and Fisher scales, are widely used but have limited discriminative capacity. This study aimed to systematically compare machine learning (ML)-based prognostic models with conventional grading systems for predicting functional outcomes and mortality after aSAH, and to evaluate factors influencing ML performance. A systematic review and meta-analysis were conducted according to PRISMA 2020 guidelines. PubMed, Embase, Scopus, Web of Science, and the Cochrane Library were searched for studies published between 2010 and 2025. Eligible studies evaluated ML-based models for outcome prediction in adult aSAH patients and reported performance of conventional grading systems. Prognostic discrimination was pooled using random-effects meta-analysis of the area under the receiver operating characteristic curve (AUC), with predefined subgroup analyses. Fourteen studies including 6,247 patients were analyzed. ML models demonstrated good to excellent discrimination, with AUCs ranging from 0.81 to 0.97. The pooled ML AUC for predicting unfavourable neurological outcome was 0.86 (95% CI 0.83-0.89; p < 0.0001), with substantial heterogeneity (I² = 96.1%). ML models outperformed conventional grading systems in most studies and showed comparable performance in the remainder. Subgroup analyses confirmed statistically significant prognostic accuracy across clinical-only, imaging-based, and multimodal ML models. Machine learning-based prognostic models demonstrate statistically significant and clinically meaningful performance for outcome prediction after aSAH, exceeding conventional grading systems and supporting their role as complementary risk stratification tools.
Spatial working memory (SWM) has been characterised as a flexible resource that determines the precision with which memoranda are stored. The 'cortical map' proposal predicts that resource allocation, and therefore mnemonic precision, is limited by the availability of cortical space to represent memoranda. This hypothesis was tested using a continuous spatial localisation task in which memory items were presented at three eccentricities and set sizes of 1-8. Localisation error increased systematically with eccentricity, as predicted by the cortical maps hypothesis. Mixture-modelling analyses indicated that the eccentricity effect was primarily driven by increases in imprecision at small set sizes (set sizes 1-5). In contrast, when set size exceeded five items, guessing made an increasingly important contribution to localisation error, suggesting that representations of peripheral locations become more vulnerable to memory failure under high memory loads. Misbinding errors were most prominent for locations nearest fixation, likely reflecting reduced inter-item spacing at small eccentricities. Stimuli were not scaled to compensate for cortical magnification, but the observed increase in imprecision closely matched predictions derived from a cortical magnification function. Together, these findings support the cortical maps hypothesis and indicate that spatial working memory representations compete for limited representational resources within the spatial maps that guide action.
 Lymph node metastasis (LNM) is a crucial prognostic indicator in perihilar cholangiocarcinoma (pCCA); however, preoperative tools for assessing LNM risk and determining the optimal extent of lymph node (LN) dissection remain limited. This study aimed to identify preoperative LNM predictors to guide regional LN dissections. This multi-institutional retrospective study included 364 patients who underwent curative-intent pCCA resection (2020-2024). Multivariate logistic regression was used to identify independent predictors of preoperative LNM. Eventually, 148 (40.7%) patients were pathologically LNM-positive. The independent preoperative predictors were age < 65 years (odds ratio [OR]: 1.634, P = 0.041), carbohydrate antigen 19-9 (CA19-9) ≥ 200 U/mL (OR: 1.868, P = 0.006), and imaging suspicion of LNM (OR: 2.863, P = 0.001). The high-risk group (imaging suspicion of LNM + CA19-9 ≥ 200 U/mL or age < 65 years) had a 66.0% pathological LNM rate, and ≥ 6 LN dissections significantly improved recurrence-free survival (RFS) compared to < 6 LN dissections (P = 0.002). The low-risk group (no risk factors) had a 23.5% pathological LNM rate, and no RFS difference was observed between the groups (P = 0.360). Preoperative imaging suspicion of LNM, elevated CA19-9 levels, and younger age can effectively identify the high-risk group with LNM in pCCA. Adequate LN dissection may be associated with improved recurrence-free survival in the high-risk group, suggesting a risk-stratified approach to the extent of LN dissection.
The prognostic nutritional index (PNI), which is calculated from the serum albumin concentrations and lymphocyte counts, predicts the outcomes in lung transplantation (LT). This study aimed to investigate the optimal timing for assessing PNI before LT. We analyzed 140 brain-dead LT recipients (2015-2023) in this study. The PNI was calculated at the time of transplant listing and immediately before LT. Using a cutoff value of 45, the patients were classified into high/low PNI groups and compared for overall survival. Changes in the PNI trajectories were analyzed. Multivariable analyses showed that a low PNI independently predicted mortality at listing (hazard ratio [HR], 2.82; 95% confidence interval [CI], 1.44-5.53; P = 0.003) and before LT (HR, 2.44; 95% CI, 1.25-4.77; P = 0.009). Patients with a persistently low PNI had a worse survival than those with a persistently high PNI (HR, 3.58; 95% CI, 1.62-7.91; P = 0.002), whereas an improvement in the PNI after listing was not significantly associated with survival. PNI at both time points significantly predicted post-transplant outcomes. A trajectory analysis highlights the importance of nutritional assessment before listing, although larger studies are required to clarify the effects of nutritional changes after listing.
Artificial intelligence (AI) is increasingly recognized as a transformative technology in healthcare, with growing evidence supporting its applicability across time-critical clinical environments. This perspective aims to evaluate the integration of AI and machine learning (ML) into pediatric emergency departments (PEDs) across three core domains: clinical decision support, stakeholder engagement, and medical education. Within clinical decision support, ML architectures have demonstrated high predictive performance across several high-acuity clinical scenarios, including triage stratification, pediatric traumatic brain injury risk classification, early sepsis detection and clinical deterioration prediction, and dermatological assessment. Model interpretability and real-world implementability remain critical prerequisites for clinical adoption, with explainability methods representing fundamental instruments to enhance transparency and stakeholder trust. Regarding stakeholder engagement, the triadic dynamic among clinicians, caregivers, and patients defines a unique communication challenge in PEDs, with large language models (LLMs) showing preliminary utility; however, stakeholder-inclusive model validation and robust data privacy protections for minors remain key challenges, particularly regarding legal ambiguities of LLM deployment in clinical pipelines. In medical education, AI-driven simulation platforms and LLM-generated adaptive curricula represent promising tools for competency-based training across pediatric emergency scenarios. Future directions emphasize the imperative of prospective multicenter validation in pediatric-specific cohorts, rigorous data quality standards addressing conformance, completeness, and plausibility, and the development of pediatric-tailored governance frameworks. Real-world implementation will require the systematic involvement of all stakeholders-including children, caregivers, clinicians, developers, and institutions-as co-designers of equitable, transparent, and safe AI systems for this uniquely vulnerable population.
Nasopharyngeal carcinoma (NPC) is a clinically aggressive malignancy in which epithelial-mesenchymal transition (EMT) plays a central role in invasion, metastasis, and therapeutic resistance. However, the coordinated transcriptional organization, regulatory mechanisms, and functional relevance of EMT-associated genes in NPC remain insufficiently defined. This study aimed to identify EMT-related hub genes in NPC and integrate multiomics and functional evidence to elucidate their biological and clinical significance. Weighted gene coexpression network analysis was performed on the GSE12452 dataset to identify EMT-associated modules and hub genes. Multiomics validation, including transcriptomic, protein level, methylation, mutation, CNV, survival, and drug response profiling, was conducted using GSCA, HPA, OncoDB, cBioPortal, and GDC databases. miRNA regulators were predicted using TargetScan and validated by dual-luciferase assays. Functional significance was assessed through RT-qPCR, Western blotting, apoptosis, viability, colony formation, and wound healing assays following treatment of NPC cells with afatinib or SB431542. WGCNA identified four EMT hub genes, VIM, SNAI1, ZEB1, and FN1, which were consistently overexpressed in NPC. These genes were hypomethylated, exhibited CNV alterations, and were associated with poor survival and broad drug resistance profiles. Predicted miRNA interactions were experimentally validated. Pharmacologic inhibition of EMT-associated signaling significantly reduced hub gene expression and suppressed malignant phenotypes across multiple NPC cell lines. This integrative analysis identifies VIM, SNAI1, ZEB1, and FN1 as a candidate EMT-associated hub signature linked with NPC progression, prognosis, and drug response patterns. Pharmacological inhibition of EMT-associated signaling reduced hub gene expression and suppressed malignant cellular phenotypes, suggesting that this EMT hub network may have potential value as a prognostic biomarker panel and as a putative therapeutic vulnerability requiring further validation.
Hyperchloremia is common in diabetic ketoacidosis (DKA), with prevalence reaching 90 %. However, its clinical impact in pediatric DKA remains unclear. This study aimed to assess the association between hyperchloremia and DKA recovery time, bicarbonate rise rate, and Pediatric Intensive Care Unit (PICU) stay, and to identify predictors of hyperchloremia. This is a retrospective cohort study that included children under 13 years admitted to the PICU with DKA between 2018 and 2024. Patients with other acute acid-base disorders were excluded. Mann-Whitney test and Chi-square test were used as appropriate. Predictors of hyperchloremia were identified using binary logistic regression. Spearman correlation was used to assess associations between chloride levels and clinical outcomes. Among 162 patients, 132 (81.5 %) had hyperchloremia and 30 (18.5 %) had normochloremia. Hyperchloremic patients had longer DKA resolution time (16 vs. 8.25 h, p<0.001), slower bicarbonate rise (0.66 vs. 0.88 mmol/L/h, p=0.004), and longer PICU stay (26 vs. 18 h, p<0.001). Chloride levels significantly correlated with DKA resolution time, bicarbonate rise, and time to reach bicarbonate ≥15 mmol/L. Hyperchloremia was independently predicted by lower body weight (aOR: 0.9, 95 % CI: 0.83-0.98, p<0.001) and use of 0.9 % NaCl as intravenous fluid (aOR: 4.94, 95 % CI: 1.41-17.4, p<0.001). Hyperchloremia in children with DKA is associated with delayed recovery and prolonged PICU stay. Lower body weight and 0.9 % NaCl use increase hyperchloremia risk. More prospective studies are needed to assess alternative fluids containing different chloride concentrations.