Browse Articles
Discover research articles across all indexed journals
A robust framework for protein-protein interaction prediction with multi-objective ensemble learning and embedding-based representations
Abstract Protein–protein interactions (PPIs) perform a key role in virtually all cellular processes. However, experimental identification of PPIs remains costly, time-consuming, and often incomplete. To address these challenges, this study presents a hybrid adaptive framework for PPI prediction that integrates modern protein language models with evolutionary optimization and ensemble learning. It uses the language model Prot-T5-XL-Uniref-50 to embed protein sequences, capturing rich contextual, structural, and physicochemical information. The resulting high-dimensional representations are then compressed using uniform manifold approximation and projection to reduce computational complexity. A hybrid approach coupling the multi-objective non-dominated sorting genetic algorithm-II (NSGA-II) with random forest is then proposed to enhance classifier robustness. This evolutionary strategy simultaneously maximizes prediction accuracy and classifier diversity while estimating the optimal number of trees required for the ensemble from the pareto-optimal fronts. Comparative results with state-of-the-art methods validate the superior performance of the proposed method across four benchmark datasets- Human , E. coli , Drosophila , and C. elegans . Finally, using SHapley Additive exPlanations, each feature’s contribution to the model’s predictions was quantified and visualized, facilitating the ranking and examination of influential embedding dimensions. Overall, the proposed framework offers a reliable and robust solution for large-scale PPI prediction based solely on protein sequence data.
The Toll-like receptor 1/2 ligand Pam3Cys inhibits memory impairment after traumatic brain injury in male and female rats
Peripheral miRNA profiling identifies therapy-specific biological trajectories in the treatment of cognitive dysfunction in depression: an exploratory mechanistic study
Abstract Cognitive dysfunction in major depressive disorder (MDD) often persists despite standard treatment. Integral cognitive remediation (INCREM) shows clinical efficacy, yet its molecular drivers remain poorly understood compared to supportive interventions like psychoeducation (PSYCHOED). In this exploratory study, we investigated whether circulating microRNAs (miRNAs) reflect the distinct biological trajectories of these interventions. We conducted a longitudinal analysis of 38 candidate miRNAs (previously validated in postmortem brain tissue of MDD individuals) in plasma from 22 patients with MDD participating in a randomized clinical trial who were assigned to 12 weeks of INCREM or PSYCHOED. Using bioinformatic pathway analyses (KEGG/GO), we identified non-overlapping miRNA profiles associated with each treatment. The INCREM response was characterized by a specific seven-miRNA signature (let-7b-3p, miR-100-5p, miR-129-5p, miR-135a-5p, miR-151a-5p, miR-4516, and miR-451a) targeting gene networks regulated in neuroplasticity, axon guidance, and synaptic transmission. These molecular changes mirrored significant objective cognitive improvements. Conversely, PSYCHOED induced a distinct profile (miR-126-5p and miR-195-5p) related to systemic cellular homeostasis and Wnt signaling, without objective cognitive gains. Our findings suggest that psychological therapies act as specific biological stimuli with distinct molecular targets. These circulating miRNA signatures provide preliminary evidence of neuroplasticity-related pathways as key drivers of cognitive recovery, offering potential blood-based biomarkers for precision psychiatry in MDD.
Sorting endosomes play key roles in presentation of Mycobacterium tuberculosis-derived ligands to MAIT cells
Abstract The immune system has developed specialized mechanisms to recognize intracellular pathogens such as Mycobacterium tuberculosis (Mtb). Major Histocompatibility Complex Class I-Related protein 1 (MR1) is a conserved nonclassical antigen presenting molecule that presents ligands derived from microbial riboflavin synthesis to Mucosal Associated Invariant T (MAIT) cells. While endosomal trafficking facilitates MR1 antigen presentation during Mtb infection, the exact mechanisms by which MR1 loading of Mtb-derived ligands occurs are not known. We found that trafficking through sorting endosomes mediates MR1 antigen presentation during Mtb infection. Sorting endosomes utilize trafficking proteins such as Syntaxin (Stx) 6, Stx12, Stx16 and VAMP4. Prior work demonstrates the importance of VAMP4 for MR1 presentation during Mtb infection; we have found that Stx12 and Stx16 are also important. Interference with Stx12 or Stx16 via siRNA-mediated knockdown reduces MR1 antigen presentation of Mtb. Using RFP-tagged constructs, we found Stx16 co-localized more with MR1 vesicles compared to Stx12 in MR1-GFP expressing airway epithelial cells. Stx12 and Stx16 blockade increase MR1 surface stabilization and total expression, indicating that impaired endosomal trafficking hinders MR1 internalization. Together, these findings support a role for sorting endosomes in the selective sampling of the intracellular environment and MR1-mediated recognition of Mtb-infected cells.
The impact of GenAI-assisted instructional design on the teaching ability of pre-service physical education teachers
A combined PIE-FRET and FCS assay to monitor RNA dynamics and cleavage by SARS-CoV-2 Nsp15
Assessment of survival and neonatal morbidity rates of very low birth weight infants in Türkiye: Turkish Neonatal Society SECRETS (TNS-SECRETS) study
Echo state network-based model reference adaptive predefined time control for magnetic levitation linear synchronous motor
Stacked multi-fusion CNN: an adaptive attention model for privacy preserving deepfake forensics
Abstract The emergence of Generative Artificial Intelligence (Gen-AI) and Generative Adversarial Network (GAN)-based deepfakes poses significant security risks in sociocultural and sociopolitical domains. This necessitates the development of advanced and effective detection methods to prevent vulnerability in social networks. Classical Machine Learning (ML) algorithms have their limitations, especially in classifying the deepfakes. To address these issues, this paper suggests a privacy-preserving Stacked Multi-Fusion (SMF) Convolutional Neural Network (CNN) approach to classify deepfakes. An improved CNN model is proposed, integrating an adaptive multi-scale attention framework with enhanced residual blocks and a Squeeze-and-Excitation (SE) mechanism. The selection of these components is backed with an ablation study to report the individual contribution to the overall optimal architecture. A hybrid lossless multilayer cryptosystem based on a chaos-based approach, Deoxyribonucleic Acid (DNA)-based computing, etc., is developed to secure images in cloud storage. The efficacy of the proposed SMF model is validated using the 140K Real and Fake Faces (RFF) image dataset. The proposed approach was found to achieve stable performance with minimal variation in various tests. It achieved a test accuracy and ROC-AUC of 97.80 and 99.79, respectively. This paper provides comprehensive relevant factors for building effective synthetic media detection systems.
Optimized ceramic powder geopolymer mortar reinforced with basalt fibers for mechanical durability and environmental performance
Predictive analysis of tensile strength in FDM Fabricated PLA wood composite materials
Pilot tests of continuous gas extraction with improved L-shaped wells from mining and goaf areas
Authorship identification for Chinese literature based on a pyramid deep bidirectional gated recurrent unit network with voting strategy
Comparative analysis of six new chloroplast genomes in Platanthera (Orchidaceae) enhances understanding of its diversification
Effects of pneumatic tube systems on next-generation viscoelastic coagulation test devices in septic patients and healthy individuals: Results of the randomized controlled VETaPT trial
Abstract Rapid coagulation assessment is essential in critical care to enable timely correction of coagulopathy. Viscoelastic testing (VET) supports this goal but may be affected by mechanical stress during transport by pneumatic tube systems (PTS). As PTS are widely used to expedite sample delivery, evaluating the robustness of next-generation VET and platelet function assays under these conditions is crucial for reliable, time-sensitive diagnostics in intensive care. This study investigated the impact of PTS transport on VET and platelet function testing in healthy individuals and septic patients, including quantitative analysis of acceleration forces. This randomized trial applied a non-systematic sample-level allocation of paired blood samples from 46 healthy volunteers and 45 septic patients to manual and PTS transport. Acceleration was quantified via three-axis accelerometry. Samples were analyzed using ClotPro, ROTEM, TEG PlateletMapping, and Multiplate. Primary objective was the difference in test results following both transport modes. Analyses were performed on paired datasets (manual vs. PTS) per participant and assay. As pre-specified in the protocol, logistic regression modeled the probability of a clinically relevant EX-test clotting time (CT) change (≥ 10 s) within each cohort. Given the absence of associations, secondary equivalence analyses (TOST [two one-sided tests] and bootstrap) assessed whether observed effects were within pre-specified bounds. Neither logistic regression nor correlation analysis indicated an effect of mechanical stress on variable changes (all ρ < 0.5; p > 0.01). Across platforms, most viscoelastic and platelet function variables remained within predefined equivalence margins after PTS transport. Exceptions were TEG PlateletMapping HKH-R and, by bootstrap, ADP/AA-inhibition. In healthy volunteers, equivalence was confirmed for all variables (TOST p < 0.001). In septic patients, minor shifts remained within clinically acceptable limits, with equivalence confirmed for ClotPro IN-test CT (± 16s, p lower = 0.036; p upper < 0.001), EX-test MCF (± 2 mm, both p < 0.001), ROTEM INTEM CT (± 16s, both p < 0.001), Multiplate TRAP (± 10U, p lower = 0.001; p upper < 0.001), and TEG PlateletMapping ADP/AA inhibition (± 5%, both p < 0.05). Most next-generation viscoelastic and platelet assays are robust to PTS-induced stress. Coagulation diagnostics can include PTS transport without compromising validity. Only selected TEG PlateletMapping variables exhibited variability, indicating limited robustness. Trial registration: The study is retrospectively registered with the German Clinical Trials Register (DRKS00036231; https://drks.de/search/de/trial/DRKS00036231/details on 20.02.2025).
Post-quantum cognitive zero trust architecture for healthcare IoT devices
Healthcare IoT systems increasingly rely on interconnected, resource-constrained devices that are vulnerable to both classical and emerging quantum-enabled cyber threats, but introduced heightened cybersecurity risks, particularly from emerging quantum computing threats that can break conventional encryption such as RSA and ECC. This study addresses the urgent need to secure resource-constrained healthcare IoT systems against both classical and post-quantum attacks while maintaining low-latency performance suitable for non-real-time clinical traffic.This study proposed the Post-Quantum Cognitive Zero-Trust Architecture (PQ-CZTA), which integrates NIST-standardized post-quantum cryptography,CRYSTALS-Kyber for key encapsulation and SPHINCS+ for stateless digital signatures,with a lightweight cognitive engine. The engine employs three machine learning classifiers (Random Forest as primary, Logistic Regression, and Multi-Layer Perceptron) trained with SMOTE oversampling and 5-fold cross-validation on six diverse intrusion detection datasets (NSL-KDD, CIC-IDS2017, MedBIoT, Edge-IIoTset, IoT-23, TON_IoT). Intrusion probabilities are converted to dynamic trust scores that drive zero-trust policy decisions (ALLOW, MONITOR, DENY, QUARANTINE) in a layered architecture enforcing least privilege and hop-by-hop re-authentication.Evaluations demonstrate excellent detection performance with F1-scores ranging from 0.972 to 1.000 across datasets, particularly strong on modern IoT traffic. The full post-quantum handshake incurs 3.1–4.4 seconds latency (dominated by SPHINCS+), which remains acceptable for periodic vital-sign reporting, alerts, and firmware updates. An ablation study proves the importance of the components, with SMOTE contributing 5–20% to the F1 score on imbalanced data and cognitive ML providing the advantage of adaptive policies over static policies.PQ-CZTA provides a practical, quantum-resilient framework that enhances patient data privacy (HIPAA compliance via adaptive risk scoring), predicts attacks on limited devices, and supports resilient IoT-enabled healthcare systems against future quantum threats.