Search Results (218)

Background: Adverse drug reactions (ADR) pose substantial risks to patient safety and challenge clinical decision-making. However, traditional predictive approaches frequently fail to deliver interpretable insights into the complex interplay between pharmaceuticals and biological systems. Methods: We propose the KRDQN (Knowledge Graph Reinforced Deep Q-Network) predictive framework. First, a knowledge graph (KG) that encompasses five entity types—drug, target, pathway, gene, and adverse drug reaction (ADR)—is constructed, and each node is enriched with intrinsic attribute features. A Deep Q-Network (DQN) is subsequently deployed within a reinforcement learning paradigm to generate interpretable ADR predictions. Model performance is evaluated by five-fold cross-validation, with accuracy and AUC reported. Finally, the Spearman correlation coefficients between drug–drug similarity and path–path similarity are computed, and case studies are conducted to further assess the predictive capability of KRDQN. Results: We evaluated KRDQN on a comprehensive data set encompassing both drug–drug interactions and ADR records. Experimental results demonstrate that KRDQN surpasses state-of-the-art baselines, attaining a recall of 0.8171 and an AUC of 0.8327. Furthermore, to demonstrate the practical value of the KRDQN prediction framework, we applied it to predict potential ADRs and their mechanism pathways for the drugs sunitinib and indomethacin. The results indicated that the KRDQN framework could identify biological mechanism pathways consistent with clinical evidence. Conclusions: In this study, we developed the reinforcement learning-based KRDQN predictive framework, which outperforms existing baselines in predictive performance and yields interpretable adverse drug reaction (ADR) predictions, thereby serving as a powerful tool for pharmacovigilance and clinical decision-making. Full article

Background: The human gut microbiota plays a central role in host physiology by influencing digestion, immune function, and metabolism. Characterizing age-associated differences in the organization of microbial metabolism may provide insights into functional variation in the gut microbiome across the human lifespan. Methods: Gut microbiota metabolic organization was analyzed in a cohort of 30 individuals spanning three age groups (infants, adults, and elderly individuals) and comprising 156 stool samples. Community metabolic networks were reconstructed using the metabolic Directed Acyclic Graph (m-DAG) framework derived from KEGG Ortholog annotations. Network topology was characterized to assess whether the resulting networks conform to previously described global structural patterns and to examine age-associated variability. Pairwise m-DAG dissimilarities were computed, and hierarchical clustering was applied to evaluate similarities among samples. Results: All samples revealed a conserved global network organization, alongside marked variability in specific structural features. Hierarchical clustering did not strictly reflect chronological age. A homogeneous cluster composed exclusively of adult samples was identified, whereas elderly samples were distributed across two clusters, one grouping with adults and the other with infants. Exploratory discriminative analyses identified functional reactions contributing to the separation between the adult cluster and the remaining samples, indicating age-associated differences in metabolic network organization. Conclusions: Gut microbiota metabolic networks in adults tend to exhibit lower redundancy and structural complexity, whereas those in infant and elderly samples display more heterogeneous network configurations. This network-based analysis provides a functional perspective on age-associated variation in gut microbiota metabolism and offers a framework for future integrative studies. Full article

Background/Objectives: Cognitive decline and compromised attention control serve as early indicators of neurodysfunction that manifest as broader psychopathological symptoms, yet conventional mental health assessment relies predominantly on subjective self-report measures lacking objectivity and temporal granularity. We propose Neurosense, an AI-driven brain health assessment framework using electroencephalography (EEG) to non-invasively capture neural dynamics. Methods: Our Dual-path Spatio-Temporal Adaptive Gated Encoder (D-STAGE) architecture processes temporal and spatial EEG features in parallel through Transformer-based and graph convolutional pathways, integrating them via adaptive gating mechanisms. We introduce a two-stage paradigm: first training on cognitive task EEG for reaction time prediction to acquire cognitive performance-related representations, then featuring parameter-efficient adapter-based transfer learning to estimate p-factor—a transdiagnostic psychopathology dimension. The adapter-based transfer achieves competitive performance using only 1.7% of parameters required for full fine-tuning. Results: The model achieves effective reaction time prediction from EEG signals. Transfer learning from cognitive tasks to mental health assessment demonstrates that cognitive efficiency representations can be adapted for p-factor prediction, outperforming direct training approaches while maintaining parameter efficiency. Conclusions: The Neurosense framework reveals hierarchical relationships between neural dynamics, cognitive efficiency, and mental health dimensions, establishing foundations for a promising computational framework for mental health assessment applications. Full article

Autism spectrum disorder (ASD) is a neurodevelopmental condition characterized by impairments in social interaction and communication, along with atypical behavioral patterns. Affected individuals often seem isolated in their inner world and exhibit particular sensory reactions. The World Health Organization has indicated a persistent increase in the global prevalence of autism, with approximately 1 in 127 persons affected worldwide. This study contributes to the growing research effort by presenting a comprehensive analysis of functional connectivity patterns for ASD prediction using rs-fMRI datasets. A novel approach was used for ASD identification using the ABIDE II dataset, based on functional networks derived from BOLD signals. The sparse functional brain connectome (Lean-NET) model is employed to construct subject-specific connectomes, from which local graph metrics are extracted to quantify regional network properties. Statistically significant features are selected using Welch’s t-test, then subjected to False Discovery Rate (FDR) correction and classified using a Support Vector Machine (SVM). Our experimental results demonstrate that locally derived graph metrics effectively discriminate ASD from typically developing (TD) subjects and achieve accuracy ranging from 70% up to 91%, highlighting the potential of graph learning approaches for functional connectivity analysis and ASD characterization. Full article

Drug–drug interactions (DDIs) can cause adverse reactions or reduce the efficiency of a drug. Using computers to predict DDIs is now critical in pharmacology, as this reduces risks, improves drug outcomes and lowers healthcare costs. Clinical trials are slow, expensive, and require a lot of effort. The use of artificial intelligence (AI), primarily in the form of machine learning (ML) and its subfield deep learning (DL), has made DDI prediction more accurate and efficient when handling large datasets from biological, chemical, and clinical domains. Many ML and DL approaches are bio-inspired, taking inspiration from natural systems, and are considered part of the broader class of biomimetic methods. This review provides a comprehensive overview of AI-based methods currently used for DDI prediction. These include classical ML algorithms, such as logistic regression (LR) and support vector machines (SVMs); advanced DL models, such as deep neural networks (DNNs) and long short-term memory networks (LSTMs); graph-based models, such as graph convolutional networks (GCNs) and graph attention networks (GATs); and ensemble techniques. The use of knowledge graphs and transformers to capture relations and meaningful data about drugs is also investigated. Additionally, emerging biomimetic approaches offer promising directions for the future in designing AI models that can emulate the complexity of pharmacological interactions. These upgrades include using genetic algorithms with LR and SVM, neuroevaluation (brain-inspired model optimization) to improve DNN and LSTM architectures, ant-colony-inspired path exploration with GCN and GAT, and immune-inspired attention mechanisms in transformer models. This manuscript reviews the typical types of data employed in DDI (pDDI) prediction studies and the evaluation methods employed, discussing the pros and cons of each. There are useful approaches outlined that reveal important points that require further research and suggest ways to improve the accuracy, usability, and understanding of DDI prediction models. Full article

Reaction conditions (RCs) are a crucial part of reaction definition, and their accurate prediction is an important component of chemical synthesis planning. The existence of multiple combinations of RCs capable of achieving the desired result complicates the task of condition recommendation. Herein, we propose a conditional variational autoencoder (CVAE) generative model to predict suitable RCs. The CVAE model has been customized to generate diverse sets of valid conditions, ensuring high flexibility and accuracy, while circumventing the necessity for enumeration or combinatorial search of potential RCs. The efficacy of the CVAE approaches was evaluated using hydrogenation reactions and other H₂-mediated reactions, predicting the set of catalysts, additives (acid, base, and catalytic poison), ranges of temperature, and pressure. The CVAE models predicted conditions with different “heads”, each corresponding to specific condition components, and their respective losses. CVAE models were tested on two datasets: a small one containing 31K reactions with 2232 potential conditions’ combinations and a big one having 196K reactions with ~7 × 10⁴² potential conditions’ combinations to evaluate the model’s ability to predict varying complexity and diversity conditions. To optimize the accuracy of the models, we experimented with three latent distribution variants—Gaussian (g-CVAE), Riemannian Normalizing Flow (rnf-CVAE), and Hyperspherical Uniform (h-CVAE). In our experiments, the h-CVAE model demonstrated robust overall performance, making it the optimal choice for scenarios requiring high accuracy across multiple top-k predictions. Benchmarking analyses demonstrated the high performance of the CVAE models compared to state-of-the-art reaction condition prediction approaches. Full article

Motivation: We aim to study the boundary stability and persistence of positive odes in mathematical epidemiology models by importing structural tools from chemical reaction networks. This is largely a review work, which attempts to congregate the fields of mathematical epidemiology (ME), and chemical reaction networks (CRNs), based on several observations. We started by observing that epidemiologic strains, defined as disjoint blocks in either the Jacobian on the infected variables, or as blocks in the next generating matrix (NGM), coincide in most of the examples we studied, with either the set of critical minimal siphons or with the set of minimal autocatalytic sets (cores) in an underlying CRN. We leveraged this to provide a definition of the disease-free equilibrium (DFE) face/infected set as the union of either all minimal siphons, or of all cores (they always coincide in our examples). Next, we provide a proposed definition of ME models, as models which have a unique boundary fixed point on the DFE face, and for which the Jacobian of the infected subnetwork admits a regular splitting, which allows defining the famous next generating matrix. We then define the interaction graph on minimal siphons (IGMS), whose vertices are minimal siphons, and whose edges indicate the existence of reactions producing species in one siphon from species in another. When this graph is acyclic, we say the model exhibits an Acyclic Minimal Siphon Decomposition (AMSD). For AMSD models whose minimal siphons partition the infection species, we show that the NGM is block triangular after permutation, which implies the classical max structure of the reproduction number $R_0$ for multi-strain models. In conclusion, using irreversible reaction networks, minimal siphons and acyclic siphon decompositions, we provide a natural bridge from CRN to ME. We implement algorithms to compute IGMS and detect AMSD in our Epid-CRN Mathematica package (which already contain modules to identify minimal siphons, criticality, drainability, self-replicability, etc.). Finally, we illustrate on several multi-strain ME examples how the block structure induced by AMSD, and the ME reproduction functions, allow expressing boundary stability and persistence conditions by comparing growth numbers to 1, as customary in ME. Note that while not addressing the general Persistence Conjecture mentioned in the title, our work provides a systematic method for deriving boundary instability conditions for a significant class of structured models. Full article

Microrobot navigation in constrained environments requires path planning methods that ensure both efficiency and collision avoidance. Conventional approaches, which typically combine graph-based path finding with geometric path simplification, effectively reduce path complexity but often generate collision-prone paths because wall boundaries are not considered during simplification. Therefore, to overcome this limitation, we present Secure Angle-based Geometric Elimination (SAGE), a single-pass path-simplification algorithm that converts pixel-level shortest paths into low-complexity trajectories suitable for real-time collision-free navigation of microrobots. SAGE inspects consecutive triplets (p_i, p_i+1, p_i+2) and removes the middle point when the turning angle is smaller than threshold (∠p_ip_i+1p_i+2 ≤ θ_th) or the direct segment (p_i → p_i+2) is collision-free. Quantitative analysis shows that SAGE achieves approximately 5% shorter path length, 20% lower turning cost and 0% collision rate, while maintaining computation comparable to the Ramer–Douglas–Peucker algorithm. Integration with Dijkstra and RRT planners confirms scalability across complex maze and vascular environments. Experimental microrobot demonstrations show navigation with complete collision avoidance, establishing SAGE as an efficient and reliable framework for high-speed microrobot navigation and automation in lab-on-a-chip, chemical-reaction and molecular-diagnostic systems. Full article

The flow of a two-dimensional incompressible hybrid nanofluid over a stretching cylinder containing microorganisms with parallel effect of inclined magnetohydrodynamic was examined in the current study in relation to chemical reactions, heat source effect, nonlinear heat radiation, and multiple convective boundaries. The main objective of this research is the optimization of heat transfer with inclined MHD and variation in different physical parameters. The governing partial differential equations are transformed into a set of ordinary differential equations by applying the appropriate similarity transformations. The Runge–Kutta method is recognized for using shooting as a technique. Surface plots, graphs, and tables have been used to illustrate how various parameters affect the local Nusselt number, mass transfer, and heat transmission. It is demonstrated that when the chemical reaction parameter rises, the concentration and motile concentration profiles drop. The least responsive is the rate of heat transfer to changes in the inclined magnetic field and most associated with changes in the Biot number and radiation parameter shown in contour plot. The streamline graph illustrates the way fluid flow is affected simultaneously by the magnetic parameter M and an angled magnetic field. Local Nusselt number and local skin friction are improved by the curvature parameter and mixed convection parameter. The contours highlight the intricate interactions between restricted magnetic field, significant radiation, and substantial convective condition factors by displaying the best heat transfer. The three-dimensional surface, scattered graph, pie chart, and residual plotting demonstrate the statistical analysis of the heat transfer. The results support their use in sophisticated energy, healthcare, and industrial systems and enhance our theoretical knowledge of hybrid nanofluid dynamics. Full article

The forecasting of Drug-Drug Interactions (DDIs) is essential in pharmacology and clinical practice to prevent adverse drug reactions. Existing approaches, often based on neural networks and knowledge graph embedding, face limitations in modeling correlations among drug features and in handling complex BioKG relations, such as one-to-many, hierarchical, and composite interactions. To address these issues, we propose Rot4Cap, a novel framework that embeds drug entity pairs and BioKG relationships into a four-dimensional vector space, enabling effective modeling of diverse mapping properties and hierarchical structures. In addition, our method integrates molecular structures and drug descriptions with BioKG entities, and it employs capsule network–based attention routing to capture feature correlations. Experiments on three benchmark BioKG datasets demonstrate that Rot4Cap outperforms state-of-the-art baselines, highlighting its effectiveness and robustness. Full article

This study focuses on the analysis of the simultaneous impact of inclined magnetohydrodynamic Carreau hybrid nanofluid flow over a stretching sheet, including microorganisms with the effects of chemical reactions in the presence and absence of slip conditions for dilatant $(n>1.0)$ and quasi-elastic hybrid nanofluid $(n<1.0)$ limitations. Meanwhile, the transfer of energy is strengthened through the employment of heat sources and bioconvection. The analysis incorporates nonlinear thermal radiation, chemical reactions, and Arrhenius activation energy effects on different profiles. Numerical simulations are conducted using the efficient Bvp5c solver. Motile concentration profiles decrease as the density slip parameter of the motile microbe and Lb increase. The Weissenberg number exhibits a distinct nature depending on the hybrid nanofluid; the velocity profile, skin friction, and Nusselt number fall when $(n>1.0)$ and increase when $(n<1.0)$. For small values of inclination, the 3D surface plot is far the surface, while it is close to the surface for higher values of inclination but has the opposite behavior for the 3D plot of the Nusselt number. A detailed numerical investigation on the effects of important parameters on the thermal, concentration, and motile profiles and the Nusselt number reveals a symmetric pattern of boundary layers at various angles $\left(\alpha \right)$. Results are presented through tables, graphs, contour plots, and streamline and surface plots, covering both shear-thinning cases $(n<1.0)$ and shear-thickening cases $(n>1.0)$. Full article

Background/Objectives: This study examined the efficacy of a physical activity (PA) intervention augmented by a non-immersive Virtual Reality (VR) gaming system (i.e., Virtual Reality-infused Movement and Activity Program; V-MAP) on physical activity (i.e., sedentary behavior, moderate-to-vigorous PA [MVPA], vigorous PA [VPA]) and cognitive skills (i.e., response error, movement latency and reaction time) in Head Start preschoolers. Methods: Using a repeated-measure with 1-month follow-up design, a sample of 13 Head Start preschoolers (Mage = 67.08 ± 4.32 months; 36.2% boys) engaged in a 6-week V-MAP intervention (30-min session; 8 sessions) that focused on non-immersive VR based movement integration. The Cambridge Neuropsychological Test Automated Battery (CANTAB) was used to measure cognition; school-based PA and sedentary behavior were assessed by ActiGraph accelerometer. Pedometers were used to monitor real time engagement and implementation over eight intervention sessions. Results: On average, children obtained 1105 steps during the 30-min intervention (36.85 steps/min). There was a significant increase in VPA after the V-MAP intervention, whereas no significant changes in MVPA or sedentary behavior were observed (ps > 0.05). Although we did not observe significant improvement in studied cognitive function variables (ps > 0.05) after the V-MAP intervention, some delayed effects were observed in the follow-up test (Cohen’s d ranges from −0.41 to −0.73). Conclusions: This efficacy trial provides preliminary support that implementing V-MAP in recess may help Head Start preschoolers achieve or accumulate the recommended daily 60-min MVPA guideline during preschool years. The findings also provide insights that VR-based PA for as little as 30 min per day may benefit cognitive capability. Full article

Path planning is a complex task in robotics, requiring an efficient and adaptive algorithm to find the shortest path in a dynamic environment. The traditional path planning methods, such as graph-based algorithms, sampling-based algorithms, reaction-based algorithms, and optimization-based algorithms, have limitations in computational efficiency, real-time adaptability, and obstacle avoidance. To address these challenges, hybrid path planning algorithms combine the strengths of multiple techniques to enhance performance. This paper includes a comprehensive review of hybrid approaches based on graph-based algorithms, sampling-based algorithms, reaction-based algorithms, and optimization-based algorithms. Also, this article discusses the advantages and limitations, supported by a comparative evaluation of computational complexity, path optimization, and finding the shortest path in a dynamic environment. Finally, we propose an AI-driven adaptive path planning approach to solve the difficulties. Full article

This paper investigates the dispersion of atmospheric pollutants in urban environments using a fractional-order convection–diffusion-reaction model with dynamic line sources associated with vehicle traffic. The model includes Caputo fractional time derivatives and Riesz fractional space derivatives to account for memory effects and non-local transport phenomena characteristic of complex urban air flows. Vehicle trajectories are generated stochastically on the road network graph using Dijkstra’s algorithm, and each moving vehicle acts as a mobile line source of pollutant emissions. To reflect the daily variability of emissions, a time-dependent modulation function determined by unknown parameters is included in the source composition. These parameters are inferred by solving an inverse problem using synthetic concentration measurements from several fixed observation points throughout the area. The study presents two main contributions. Firstly, a detailed numerical analysis of how fractional derivatives affect pollutant dispersion under realistic time-varying mobile source conditions, and secondly, an evaluation of the performance of the proposed parameter estimation method for reconstructing time-varying emission rates. The results show that fractional-order models provide increased flexibility for representing anomalous transport and retention effects, and the proposed method allows for reliable recovery of emission dynamics from sparse measurements. Full article

Reduced chemical kinetic mechanisms are essential for enabling the use of complex fuels in 3D CFD combustion simulations. This study presents the development and optimization of a compact mechanism capable of accurately modeling ethanol–gasoline blends, including Brazilian Type-C gasoline (27% ethanol by volume) and up to pure ethanol (E100). An initial mechanism was constructed using the Directed Relation Graph with Error Propagation (DRGEP) method applied to detailed mechanisms selected for each surrogate component. The resulting mechanism was then refined through three global iterations of a genetic algorithm targeting ignition delay time (IDT) and laminar flame speed (LFS) performance. Five candidate versions (Mec1 to Mec5), each containing 179 species and 771 reactions, were generated. Mec4 was identified as the optimal configuration based on quantitative error analysis across all tested conditions and blend ratios. The final mechanism offers a balance between predictive accuracy and computational feasibility, making it well-suited for high-fidelity simulations in complex geometries involving multi-component ethanol–gasoline fuels. Full article