Background: Accurate prognosis in diffuse large B-cell lymphoma (DLBCL) is limited by biological heterogeneity and the absence of formal per-patient uncertainty quantification for treatment-response prediction. This study introduces a multi-layer evidence fusion framework combining gene expression profiling and clinical features with distribution-free uncertainty quantification. Methods: The proposed framework integrates four evidence layers—WGCNA co-expression eigengenes, ssGSEA pathway scores, bootstrap-stable prognostic genes, and the International Prognostic Index—through AUC-proportional reliability discounting and sequential Dempster–Shafer fusion. The primary endpoint was three-year overall survival (OS3yr) as a surrogate for R-CHOP treatment response. Inductive conformal prediction (ICP, ε = 0.10) was applied to provide per-patient uncertainty sets with a distribution-free coverage guarantee. Training used GSE10846 (n = 223, Affymetrix); external validation used GSE181063 (n = 479, Illumina). Results: The proposed framework achieved internal AUC = 0.808 (95% CI [0.750, 0.863]), significantly outperforming logistic stacking (AUC = 0.786, p = 0.0009) and unweighted DS fusion (AUC = 0.767, p = 0.037). External AUC = 0.791 was statistically comparable to logistic stacking (DeLong p = 0.21). AUC-proportional discounting reduced inter-source conflict K- by 75% (0.093→0.023). ICP achieved 90.1% internal and 94.6% external coverage; 43.5% of training patients received uncertain predictions ({S,R}). Conclusions: The proposed framework provides an uncertainty-aware approach for multi-layer genomic–clinical evidence fusion in DLBCL, with cross-platform discrimination validated on an independent Illumina cohort.
Background: Near-perfect machine-learning performance on public clinical datasets can arise from data leakage or predictors unavailable at the intended prediction time. We audited a breast cancer vital-status workflow and established a censoring-aware benchmark. Methods: The public SEER-derived dataset contained 4024 women diagnosed in 2006–2010. After removing one duplicate, 616 of 4023 records had status “Dead”. We reproduced a support-vector machine workflow that oversampled the complete dataset before an 80:20 split, quantified exact train-test row overlap, and repeated the analysis with training-only oversampling. A baseline-only support-vector machine and a penalized Weibull accelerated failure-time model were then evaluated using training-only tuning and an untouched test set. Results: Pre-split oversampling placed duplicate feature rows from 50.5% of test observations in training and yielded an area under the receiver operating characteristic curve (AUROC) of 0.998. Splitting first eliminated overlap and reduced AUROC to 0.715 under the retained legacy model. The tuned baseline-only model achieved AUROC 0.745 (95% confidence interval 0.696–0.793). The survival model achieved a test concordance index of 0.753 (0.708–0.797). Conclusions: Resampling order and post-baseline information materially inflated apparent performance. The corrected results support moderate internal prognostic discrimination, not breast cancer diagnosis, triple-negative subtype classification, or clinical deployment.
Background: Deep learning models, particularly convolutional neural networks (CNNs), have shown promising performance for pneumonia detection using chest X-ray images. However, the impact of preprocessing, architecture selection, data augmentation, and ensemble strategies has not been systematically evaluated. This study investigated how these factors affect model robustness and diagnostic performance. Methods: A public pediatric chest X-ray dataset was used to systematically evaluate pixel normalization methods, six CNN architectures, progressive data augmentation strategies for class imbalance, and both feature-level and decision-level ensemble approaches. Model performance was assessed by considering not only overall classification accuracy but also clinically relevant risk metrics, particularly false-negative rates. Results: Pixel normalization to the 0–1 range improved model convergence, while Xception and InceptionV3 achieved the best overall performance. Model-specific augmentation strategies were more effective than a fixed 1:1 class ratio for reducing false negatives. Feature-level ensembles tended to overfit, whereas decision-level ensembles provided more stable but only modest performance improvements. Conclusions: These findings demonstrate that reliable medical AI systems require systematic optimization of preprocessing techniques, model architecture, data augmentation strategies, and clinically meaningful evaluation metrics rather than maximizing a single performance indicator. The proposed framework provides practical guidelines for developing robust deep learning models for pneumonia diagnosis in clinical settings.
Digital health observatories require predictive frameworks that transform administrative health data into reliable territorial risk indicators. We developed a biomedical informatics framework for machine-learning-based epidemiological risk stratification in the Colombian Orinoquía, with explicit attention to data-leakage prevention, validated population denominators, and unit-of-analysis discipline. From 354,088 morbidity records (2018–2023; Arauca, Casanare, Meta, Vichada) we derived 20,212 independent strata (municipality × year × diagnostic group × sex × age category × health component) and computed morbidity rates using official population projections from the Colombian National Administrative Department of Statistics (DANE), correcting a denominator instability present in the original extract. After excluding leakage-generating variables, Gradient Boosting, Random Forest, and a one-hot Logistic Regression baseline were evaluated through temporal validation, group-aware cross-validation, leave-one-department-out validation, and a geographic ablation experiment. Under leakage-controlled, stratum-level conditions, Gradient Boosting achieved an area under the receiver operating characteristic curve (AUC-ROC) of 0.913 [95% confidence interval (CI): 0.903–0.923] (Brier: 0.106). SHapley Additive exPlanations (SHAP) analysis identified diagnostic group and age category as the dominant predictors, supported by broadly stable performance without geographic identifiers (AUC 0.871) and consistent cross-department transferability (0.807–0.889); a residual municipality-level clustering effect is reported transparently. Validated denominators reversed the apparent territorial gradient, with the most remote department exhibiting the lowest documented morbidity, consistent with under-registration.
Diabetic foot ulcers (DFUs) are a major cause of lower-extremity amputation, and image-based decision support may assist screening and preliminary triage where specialist access is limited. We propose DFU-MambaKAN, a hybrid architecture combining a pure-PyTorch selective state-space (Mamba) block with a Gaussian radial-basis-function Kolmogorov–Arnold Network (RBF-KAN) feed-forward layer. The study evaluates two distinct tasks: (1) binary normal-versus-ulcer screening and (2) a dataset-specific four-class Wagner-Meggitt severity-grading task. DFU-MambaKAN contains 1.062 million parameters and was compared with ResNet50, EfficientNet-B0, MobileNetV3-Small, and ViT-Tiny under the manuscript’s common downstream training protocol. The reported values are point estimates from one deterministic 70/15/15 split generated with seed 42; no confidence intervals, repeated-seed averages, or inferential significance tests were obtained. On binary screening, DFU-MambaKAN achieved 97.17% accuracy, macro-F1 0.970, and AUC 0.994. On four-class grading, it achieved 67.75% accuracy, macro-F1 0.677, and AUC 0.895, whereas the baselines achieved 98.14–99.00% accuracy. This large gap means that the present evidence does not establish competitiveness for multiclass severity grading. Possible contributors include training duration and convergence, dataset size and class definitions, label or image-quality uncertainty, hyperparameter selection, global token mixing, and architecture–task mismatch; extended learning-curve analysis and multi-seed evaluation are needed to test these explanations. An automated technical audit identified 342 exact duplicates among 1055 class-folder images in Dataset A (32.4%); no deduplicated-versus-nondeduplicated performance comparison was performed, so this finding is reported as a methodological caution rather than proof of accuracy inflation. The model’s 1.062M parameter count supports parameter-efficient storage, but its measured single-image GPU latency was 90.54–93.14 ms, substantially higher than the baselines. Accordingly, the current evidence supports further investigation for batched or queued screening more strongly than immediate real-time mobile deployment. External clinical validation, multi-seed analysis, statistical testing, convergence studies, and component ablations remain necessary.
Accurate prediction of drug–target interactions (DTIs) is a cornerstone of computational drug discovery, with the potential to reduce experimental costs and accelerate therapeutic development. Although deep learning methods such as graph neural networks and Transformers have achieved notable success in predicting DTIs, they often fail to fully exploit the rich semantic information contained in textual descriptions of drugs and targets. We developed DrugLM, a unified framework that integrates embeddings from large language models (LLMs) into DTI prediction architectures. Textual descriptions of drugs and targets were encoded using pre-trained LLMs, with optional selective upper-layer fine-tuning for task adaptation. The resulting embeddings were incorporated into six representative backbone models and evaluated on a curated DrugBank-derived dataset. Incorporating LLM-derived embeddings improved prediction performance across the evaluated architectures. Even without task-specific adaptation, pre-trained embeddings outperformed random or conventional initialization, while selective fine-tuning yielded additional gains in many settings. A simple multilayer perceptron using only LLM-derived embeddings also surpassed several established DTI methods, highlighting the predictive value of semantic representations. DrugLM provides an architecture-flexible strategy for incorporating semantic information from textual descriptions into the evaluated DTI prediction pipelines. These findings motivate further evaluation on additional datasets and protocols.
Dermoscopic skin lesion classification is a task of major clinical importance but is computationally expensive, making it inaccessible in resource-constrained healthcare settings. In this paper, we introduce a computationally efficient skin lesion classification framework for seven classes using EfficientNet-B0, complemented by Monte Carlo (MC) Dropout for uncertainty quantification. Our approach was trained and tested on the HAM10000 dataset containing 10,015 dermoscopic images across seven classes. To address the severe 67:1 class imbalance, we employ WeightedRandomSamplerand class-weighted cross-entropy loss as complementary corrections acting at the batch-composition level and the gradient-magnitude level respectively. By performing T=50 stochastic forward passes during inference, we decompose predictive uncertainty into aleatoric and epistemic components and apply an entropy-based referral threshold that flags uncertain predictions for specialist review. To validate spatial interpretability, Gradient-weighted Class Activation Mapping (Grad-CAM) is applied and quantitatively evaluated via Intersection over Union (IoU) against ISIC segmentation masks, yielding a mean IoU of 0.61 across all accepted predictions. Our experiments achieve a test macro AUROC of 0.9404and macro F1-score of 0.7308, with six of seven classes exceeding 70% per-class accuracy (melanocytic nevi: 69.8%). Referring the 30% most uncertain predictions to a clinician raises accepted-subset AUROC from 0.9404 to 0.9568 (+1.64%). The framework is competitive with ResNet-50 and DenseNet-121 at one-fifth the parameter count, and the only lightweight method in the comparison providing calibrated uncertainty estimates. Inference latency benchmarks on an NVIDIA Jetson Nano (edge CPU mode) are reported to contextualize deployment feasibility.
The LW-9 peptide was previously identified through screening of purified molecules from the venom of the spider Phoneutria nigriventer and has been described as an immunomodulatory compound with potential application in cancer therapy. The molecule was biochemically characterized and subjected to in silico analyses to evaluate its toxicity and efficacy profiles. LW-9 exhibits a molecular mass of 1235.522 Da and the amino acid sequence PyrKKDRFLGLM-CONH2. Secondary and tertiary structures were modeled using computational approaches. The results indicated low permeability across biological membranes, including the blood–brain barrier. In silico predictions further suggested that LW-9 is neither carcinogenic nor mutagenic; however, it may act as a potent inhibitor of the CYP3A4 enzyme, raising concerns regarding potential drug–drug interactions. Overall, LW-9 shows promise as an immunomodulatory agent for cancer applications, while its pharmacokinetic properties and possible interactions with other drugs continue to be investigated.
Antiretroviral drugs (ARVDs) remain the cornerstone of HIV/AIDS management, but their therapeutic efficacy and safety are highly influenced by bioconversion processes such as hepatic metabolism and enzymatic transformation. Variability in metabolic pathways, mediated by cytochrome P450 enzymes and other liver-based systems, contributes to interindividual differences in drug response, toxicity, and resistance. Recent advances in artificial intelligence, particularly artificial neural networks (ANNs), offer promising tools for modeling and optimizing these complex bioconversion processes. ANNs are capable of learning nonlinear relationships from high-dimensional datasets, making them ideal for predicting the pharmacokinetic parameters, enzyme–substrate interactions, and metabolic stability of ARVDs. This review explores the emerging role of ANNs in understanding and optimizing the metabolic transformation of antiretroviral agents. Key applications are discussed, including prediction of drug–enzyme interactions, in silico modeling of hepatic clearance, and simulation of enzyme kinetics. The integration of molecular descriptors, omics data, and clinical parameters into ANN models allows for improved prediction accuracy and personalized therapy. Furthermore, ANN-based tools can aid in early-stage drug development by identifying metabolic liabilities and guiding structural modifications to enhance metabolic stability. Despite their potential, challenges such as data scarcity, model interpretability, and standardization remain. Future research should focus on hybrid models combining ANN with mechanistic pharmacokinetics, the incorporation of real-world patient data, and validation against experimental outcomes. Overall, ANNs represent a powerful approach to optimizing ARVDs bioconversion, with the potential to improve efficacy, reduce toxicity, and support the development of next-generation antiretroviral therapies
Background: Targeting Apolipoprotein E4 (ApoE4) represents a frontier in Alzheimer's disease therapeutics. This study investigates the therapeutic potential of a nutraceutical panel (Polydatin, trans-resveratrol, luteolin, and PEA) by exploring their interaction with the ApoE4 EZ-482 cavity. Methods: Using a dual-platform docking strategy (SwissDock and Schr & ouml;dinger Maestro) across three structural constructs. Results and Discussion: We identified the full-length protein (1-299) as the optimal target, showing a robust correlation between normalized docking scores (Spearman rho = 0.79). Crucially, biophysical analysis via dynamic light scattering (DLS) revealed that the ApoE4-oxLDL complex exhibits a zeta-potential of -10.97 mV, a state prone to pathological aggregation. Luteolin and PEA effectively altered this electrostatic environment, inducing significant positive shifts to +2.15 mV and +1.05 mV, respectively. The alignment between computational rankings and experimental zeta-potential perturbations supports the predictive reliability of our model. These findings suggest that nutraceuticals can modulate the ApoE4-oxLDL biophysical profile and highlight that a full structural context is mandatory for developing effective ApoE4-targeted interventions.
Lung cancer is the deadliest cause of cancer-related deaths worldwide, and early and accurate detection is key to improving patient outcomes. IQ-OTH/NCCD CT scan images are used in this study to present an optimized computer-aided diagnosis (CAD) framework for lung cancer detection. In order to extract deep features and improve diagnostic accuracy, a weighted geometric mean (WGM) ensemble of pretrained convolutional neural networks (CNNs) called the LCD-VRD model—comprising VGG16, ResNet50V2, and DenseNet121—provides robust feature extraction and strong generalization capabilities for accurately classifying normal, benign, and malignant (cancerous) cases. To actively mitigate data imbalance and reduce model overfitting, real-time data augmentation alongside rigorous class weighting was implemented. The results show that, with 97.27% accuracy and a 97.24% F1-score, the WGM ensemble of these models performs exceptionally well. Furthermore, Gradient-weighted Class Activation Mapping (Grad-CAM) visualization was investigated on CT images to provide an exploratory qualitative visualization of the image regions associated with model predictions. While the proposed framework shows promise as an effective tool for automated lung cancer diagnosis, its validation is currently limited to the IQ-OTH/NCCD dataset. External dataset evaluation will be essential to fully establish robustness and clinical applicability.
Myocardial edema is associated with cardiac electrical instability, but the cellular mechanisms linking osmotic cell swelling to arrhythmias remain unclear. Hypoosmotic conditions are hypothesized to drive transitions between dynamical regimes (e.g., spiral waves and multiple wavelets), producing distinct calcium oscillatory dynamics that act as markers of the underlying electrophysiological state. This study presents an integrated computational framework combining analysis of optical mapping data with mechanistic mathematical modeling to investigate calcium dynamics in cardiomyocyte monolayers under varying extracellular osmolality conditions. We developed an enhanced signal processing pipeline that reconstructs dynamic baselines from local minima using piecewise linear interpolation, enabling robust detection and characterization of calcium transients in highly heterogeneous and aperiodic signals. The computational workflow incorporated peak detection algorithms adapted for irregular oscillatory patterns, extraction of calcium transient features (amplitude, time to peak, decay durations at 30%, 50%, and 80% of peak amplitude) across spatial regions corresponding to different excitation regimes, and mathematical modeling to investigate the effects of hypoosmotic swelling at a cellular level. The parameters of the Gattoni (2016) rat ventricular cardiomyocyte model were modified to match experimental observations of the calcium transients. Simulation suggests that hypoosmotic swelling increases sarcolemmal calcium pump activity and elevates cytosolic concentrations of calmodulin and troponin, promoting alternans and delayed afterdepolarizations.
The increasing prevalence of autoimmune thyroid diseases and thyroid cancer highlights the urgent need for improved diagnostic support approaches. Traditional diagnostic methods often rely primarily on biochemical markers or qualitative imaging evaluations, which may delay accurate disease identification and hinder timely treatment. The present study demonstrates that machine learning models integrating biochemical, demographic, and ultrasound data achieve strong classification performance for thyroid disorder identification. Tree-based algorithms, such as XGBoost and Random Forest, demonstrated strong performance, while deep learning models achieved high accuracy in imaging-based classification tasks. Although the results highlight the potential of multi-source data-driven approaches to support clinical decision-making, performance variability indicates the need for validation on larger and more diverse datasets. Future work should focus on expanding data sources, incorporating additional biomarkers, and improving model interpretability to facilitate clinical translation.
Background: Hemagglutinin (HA) is the primary surface protein of the influenza A virus, determining its subtype and antigenic properties. Traditional subtype classification methods rely on DNA or amino acid sequence analysis, which does not account for protein spatial folding. Methods: In this work, we propose EpitopeGNN-a graph neural network (GNN) that constructs a residue interaction network (RIN) from the 3D structure of HA and classifies the virus subtype. The model was trained on 249 structures from the Protein Data Bank (PDB), containing H1N1, H3N2, H5N1, and other subtypes. Results: After rigorous sequence redundancy reduction (92% identity), the model maintained 95-100% accuracy on non-redundant data, significantly outperforming sequence-only baselines (the best baseline achieved 85% for multi-class and 92.3% for binary classification). A significant correlation was found between the obtained structural embeddings and phylogenetic distances (r = 0.38, p < 0.001), confirming their biological relevance and opening opportunities for structural monitoring of virus evolution, as well as rapid analog searching for novel strains. Conclusions: We developed a new graph neural network that classifies influenza A virus subtypes directly from the 3D structure of hemagglutinin using residue interaction networks and physicochemical features, which can serve as a foundation for predicting influenza virus receptor specificity and epitope immunogenicity.
Heterogeneity is central to the credibility, transportability, and clinical interpretation of meta-analytic evidence, yet its assessment is frequently reduced to isolated or misinterpreted statistics. This tutorial provides a clinically oriented framework for interpreting heterogeneity across therapeutic, diagnostic, and prognostic evidence synthesis. It distinguishes sampling error from genuine between-study variability and explains why Q, I-2, tau(2), and prediction intervals should be interpreted as complementary rather than interchangeable measures. Rather than relying on I-2 alone, meaningful interpretation requires assessing both the relative and absolute scale of between-study variability and its implications for future clinical settings. The tutorial also addresses estimator dependence, uncertainty in tau(2), the limitations of different confidence-interval approaches in random-effects models, and the clinical interpretation of wide prediction intervals, including links with certainty-of-evidence judgments. Diagnostic test accuracy meta-analysis is discussed as a bivariate problem that requires hierarchical models, variance components, correlation parameters, and prediction regions, rather than univariate I-2 summaries. Prognostic reviews are framed around case mix, baseline risk, discrimination, calibration, and transportability. Worked simulated examples, formal expressions, and practical algorithms are used to support decision-making about when pooling is reasonable, when it should be qualified, and when it should be avoided. Accurate interpretation of heterogeneity requires domain-appropriate modeling, transparent reporting, and clinical judgment rather than reliance on any single statistic.
This study evaluates a lightweight, edge-deployable artificial intelligence pipeline to assist, not replace, trained human readers in the classification of RPR test reactions. Two separate and non-directly comparable experimental configurations were investigated: a binary task (Reactive vs. Non-Reactive) using 243 original images and a ternary task (Reactive, Minimally Reactive, Non-Reactive) using a distinct dataset of 293 original images. Because the datasets were acquired using a single device and laboratory protocol, and because deterministic augmentation generates highly correlated transformations rather than independent clinical samples, the reported results should be interpreted as preliminary internal evidence of feasibility rather than proof of clinical generalizability. In the augmented internal test evaluation, the binary model achieved 99.98% accuracy (25,137/25,200), while the ternary model achieved 91.12% accuracy (14,417/15,822). In the original-image deployment evaluation, binary performance remained 100% (58/58) across FP32, FP16, and INT8; ternary performance was preserved under FP32/FP16 at 95.24% (80/84) but decreased to 76.19% (64/84) after INT8 quantization. An additional stochastic augmentation experiment for ternary INT8 deployment restored performance to 95.24% (80/84) and 0.9444 Macro-F1, but external validation remains mandatory before any clinical adoption.
Background: This study examines the clinical efficacy of daratumumab when combined with other therapeutic agents in patients with multiple myeloma, with a focus on key outcomes including overall response, progression-free survival (PFS), and stringent complete response (sCR). Methods: A systematic literature search was conducted using PubMed, Web of Science, Scopus, and the Cochrane Library. Statistical analyses were performed with R software (version 4.2.2). Between-study heterogeneity was assessed using the Cochrane Q test and the I2 statistic, while potential publication bias was evaluated through Egger’s regression test and visual inspection of funnel plots. Results: The meta-analysis revealed that daratumumab-containing regimens were associated with a 54.4% reduction in the risk of disease progression or death (hazard ratio[HR] 0.4558; 95% confidence-interval [CI]: 0.4031–0.5154). Similar results were observed using a random-effects model (HR 0.4667; 95% CI: 0.3771–0.5776), despite moderate heterogeneity (I2 = 66.7%). Moreover, patients treated with daratumumab were approximately 2.4 times more likely to achieve a stringent complete response (odds-ratio[OR] 2.38; 95% CI: 1.80–3.15), with moderate heterogeneity across studies (I2 = 58.2%). Conclusions: Incorporating daratumumab into standard therapy for multiple myeloma significantly enhances progression-free survival and the rate of stringent complete response. Despite some heterogeneity, the consistent positive outcomes support its use as an effective treatment option in clinical practice.
Background: Liver fibrosis arises from chronic liver injury and remains a major clinical challenge due to its progression toward cirrhosis and hepatocellular carcinoma, as well as the absence of approved antifibrotic therapies. This study aimed to characterize the transcriptomic behavior of the EGR1-FOS axis in liver fibrosis and its evolution into hepatocellular carcinoma, and to identify genes shared between liver fibrosis and cirrhosis. Methods: An integrated bioinformatics approach was applied to GEO transcriptomic datasets. Differentially expressed genes in hepatic fibrosis were identified using GSE139602, GSE84044, and GSE49541, with GSE62232 as control when needed, while GSE14323 and GSE89377 were used to detect genes common with cirrhosis. GEPIA, TIMER, and TISCH2 were used to assess the involvement of the EGR1-FOS axis in hepatocellular carcinoma. External validation of EGR1 expression dynamics and its coregulation with FOS was performed using the GSE135251 dataset. Results: Eleven hub genes were identified, with emphasis on the EGR1-FOS axis. EGR1 expression fluctuated across liver fibrosis etiologies, whereas FOS was predominantly downregulated. A strong correlation between EGR1 and FOS (r = 0.77) was observed, remaining stable across fibrosis stages (all p < 0.001) and in hepatocellular carcinoma (r = 0.698, p = 1.81 & times; 10(-55)). Despite overall downregulation, both genes increased progressively with advancing fibrosis (EGR1: p = 0.0008-0.0035; FOS: p = 0.0001-0.0188). Four genes were shared between fibrosis and cirrhosis (SOX9, CD24, CXCR4, and CYP2C19). Conclusions: The EGR1-FOS axis acts as a dynamic regulator of liver fibrosis and its progression, and both this axis and the four shared genes identified may serve as valuable biomarkers and potential therapeutic targets.
Sepsis is responsible for approximately 270,000 deaths annually in the United States. Conventional scoring systems, such as SOFA and qSOFA, are largely reactive and do not effectively leverage longitudinal ICU data for early prediction. This study aims to develop a deep learning framework capable of predicting sepsis onset up to 6 h before Sepsis-3 criteria are met while also providing clinically interpretable temporal explanations. The PhysioNet/CinC 2019 Challenge dataset, comprising 1,552,210 patient hours from 40,336 ICU patients, was utilized. A temporal transformer encoder (TTE) was trained using 12-h look-back windows with 92 engineered features. Severe class imbalance (2.6% positive rate) was addressed through weighted random sampling and focal loss. Fivefold patient-level cross-validation was employed to prevent temporal leakage. Platt scaling was applied for probability calibration. Grad-CAM was adapted for temporal explainability, while SHAP was used for feature-level attribution. BiLSTM-Attention and XGBoost models served as baseline comparators. The TTE model achieved a cross-validated AUROC of 0.8320±0.0032 and an AUPRC of 0.1505±0.0148, significantly outperforming BiLSTM-Attention (AUROC: 0.7859) and XGBoost (AUROC: 0.7731; DeLong p<0.0001). Platt scaling reduced the expected calibration error from 0.3154 to 0.0017. The median alert lead time was 46.5 h (IQR: 21–84 h), with 95.3% of septic patients receiving alerts at least 3 h before onset. Grad-CAM analysis identified time steps t−10 and t−9 as the most predictive. However, high-severity patients (SOFA proxy ≥ 3) demonstrated substantially reduced performance (AUROC: 0.257). The proposed TTE framework demonstrated strong and well-calibrated early sepsis prediction with substantial clinical lead time. The concentration of predictive signals 10–11 h prior to alert generation supports the feasibility of continuous automated ICU monitoring from admission onward. Reduced performance in high-severity patients highlights the need for severity-stratified modelling in future research.
Gliomas represent the most prevalent type of brain tumor, with their most aggressive variant, glioblastoma multiforme, associated with high mortality rates. Due to their elevated molecular heterogeneity, accurate classification of gliomas has presented significant challenges. Therefore, considerable effort has been dedicated to identifying relevant biomarkers that improve early diagnosis and unveil new areas for treatment. Advances in high-throughput sequencing technology have enabled public resources such as The Cancer Genome Atlas (TCGA) to provide large-scale data from various cancers, allowing researchers to perform more comprehensive analysis of this disease. In this study, we introduce MOHVAE-B, a comprehensive framework designed for the integration of multi-omics data and biomarker discovery using data from TCGA. MOHVAE-B employs a supervised hierarchical variational autoencoder integrated with SHAP-based interpretability to effectively integrate high-dimensional multi-omics data and extract the most influential features driving the model’s predictions. Subsequently, Bayesian Networks (BNs) are constructed to model conditional dependencies between the selected features, providing insights into their possible relations. Applied to the TCGA glioma cohorts, MOHVAE-B achieved a near-perfect AUC of 0.9993 and successfully identified high-impact features related to glioma classification. For glioblastoma multiforme, this included six novel candidates: LINC02172, NACA2, LINC01114, HNRNPA1P48, PPIAL4G, and LINC01558. For low-grade gliomas, the model highlighted AMER2 as a promising marker. Across both cohorts, PMP2 stood out as a particularly strong candidate for a potential role in glioma pathogenesis. The constructed BNs provided an additional layer of validation, reinforcing NACA2 as a candidate of interest in glioma biology.
