Research Appraisals
Evidence-based critical appraisals of the latest medical research, systematically evaluated using Oxford CEBM methodology.
Showing 32 appraisals
Journal of computer-aided molecular design
Artificial intelligence in impurity prediction: current landscape, challenges, and future directions
Earlier identification, regulation of impurities in pharmaceutical products is critical throughout the medication development process because they have a high impact on drug quality, safety, and regulatory approval. Traditional analytical methods like LC-MS and NMR are frequently used, but they are labour intensive, time consuming and have limited capacity to detect unknown or trace level impurities. New opportunities to forecast impurity formation before experimental observation have been enabled by developments in cheminformatics and computer modelling. This study covers advanced chemical representation techniques, data generation, data curation, and model validation with highlighting recent advances in machine learning and deep learning algorithms for impurity prediction. In association with forecasting synthetic by products, degradation products, and harmful contaminants, traditional QSAR techniques, and contemporary deep learning models such as graph neural networks, transformer based architectures, and generative frameworks are examined. Additionally, discussed the increasing significance of explainable models for regulatory body acceptability. Lastly, newly developed advanced techniques like digital twins, automated impurity profiling, integrated reaction degradation modelling, and real-time monitoring are examined, demonstrating how computational methods are transforming impurity assessment from a reactive effort into a predictive and preventive procedure.
23 July 2026
Read appraisal →Environmental science & technology
Bridging the Gap between External and Internal Exposure: Challenges and a Machine Learning Approach Validated Using Benzene Ring-Containing Pollutants
Establishing quantitative linkages between external and internal exposures is essential for accurate exposure and health risk assessment, yet this goal remains hindered by multiple challenges. This Perspective systematically reviews major challenges limiting the quantitative linkage across three dimensions: (i) External exposure: complex multimedia and multipathway exposure, environmental factor variability, and substantial data gaps; (ii) Internal exposure: difficulties in biological sample collection, large-scale biomonitoring, and appropriate biomarker selection; and (iii) External-internal exposure linkage: interindividual variability, scarce toxicokinetic parameters, and methodological limitations. In this context, the Perspective proposes ML-EIExpLink (machine learning-based external-internal exposure linkage), a framework integrating multisource, multidimensional variables, including pollutant physicochemical properties, ADME parameters, exposure factors, environmental conditions, and socioeconomic indicators, to model complex external-internal exposure relationships. An exploratory case study on benzene ring-containing pollutants demonstrates the framework's feasibility and reliability. Using gradient boosting decision trees as the core predictive engine, the model achieved robust predictive performance, with an external validation Qext2 of 0.784. Most key variables, including external air concentration, molecular planarity, and temperature, exhibited nonlinear and threshold effects on predicted blood concentrations. This Perspective introduces a novel ML-based framework for establishing robust, interpretable external-internal exposure linkages, thereby providing a useful basis for improved exposure assessment and health risk evaluation.
22 July 2026
Read appraisal →Critical care explorations
Drug Clearance and Dosing During CytoSorb Hemoadsorption: A Systematic Review
OBJECTIVE: CytoSorb (CS) hemoadsorption is increasingly used in critical care and cardiac surgery to remove inflammatory mediators and select drugs. This systematic review aimed to summarize available evidence on intentional drug removal as well as unintentional drug removal during CS therapy. DATA SOURCES: PubMed, congress abstracts, and CytoSorbents online literature database. STUDY SELECTION: Eligible studies included benchtop, animal, and human pharmacokinetic investigations of drug removal during CS therapy as well as clinical reports describing CS use for drug overdose. DATA EXTRACTION: Data from the eligible studies were extracted independently by two authors into Microsoft Excel. DATA SYNTHESIS: Data were identified for drugs across multiple therapeutic classes, including analgesics, antiarrhythmics, anticonvulsants, antidepressants, antihypertensives, anti-infectives, antithrombotics, anxiolytics, and immunosuppressants. Drugs were categorized into low (< 30%), moderate (30-60%), or high (> 60%) removal rate, or by the relative increase in clearance attributable to CS: negligible (< 25%), low (25-100%), moderate (100-400%), or high (> 400%). Based on the assembled data and considering drug-, patient-, and circuit-specific factors, general dosing recommendations for concomitant medications were developed. CONCLUSIONS: CS therapy can alter drug concentrations for select agents with variable clinical relevance. CS can reduce perioperative bleeding risk through antithrombotic removal and treat certain drug overdoses. Unintentional drug removal during CS therapy may occur and practical guidance is provided for drugs with known susceptibility to removal.
16 July 2026
Read appraisal →Nucleic acids research
DeepCYP: an integrated deep learning web server for the holistic 'pathway-site product' prediction of CYP450 metabolism
CYP450 (cytochrome P450)-mediated drug metabolism is a critical determinant of pharmacokinetics and clinical safety, making comprehensive metabolic profiling essential for rational drug discovery. Here, we present DeepCYP (https://deepcyp.scbdd.com), a freely accessible deep-learning web server for end-to-end CYP450 metabolic profiling. Trained on an expanded dataset and a mechanism-based reaction rule library, DeepCYP uses a multi-task graph neural network (GNN) combined with multi-scale descriptors. Operating directly on 2D molecular graphs, this architecture bridges the entire "pathway-site-product" continuum across nine major CYP isoforms (CYP1A2, CYP2A6, CYP2B6, CYP2C8, CYP2C9, CYP2C19, CYP2D6, CYP2E1, and CYP3A4) within a unified pipeline. Benchmarking demonstrates that DeepCYP outperforms established tools, including FAME3, SMARTCyp, and BioTransformer 3.0, improving Top-1 and Top-2 ranking metrics by over 10%. Furthermore, the server supports high-throughput batch processing, capable of evaluating ~280 molecules per minute. DeepCYP also enhances interpretability through an interactive visualization interface, featuring susceptibility radar charts and dynamic transformation tables. By translating abstract predictions into biological insights, DeepCYP provides a practical tool to accelerate lead optimization and mitigate toxicity risks.
13 July 2026
Read appraisal →Journal of hazardous materials
Biomarker-anchored screening of dietary microplastics via a gut microbiota-informed machine learning model
Microplastics (MPs) are increasingly recognized as emerging contaminants in the human diet, yet the absence of unified biomarker-anchored screening thresholds hampers quantitative risk assessment. This study establishes a gut microbiota-anchored machine learning framework that links Firmicutes/Bacteroidetes (F/B) ratio thresholds with real-world dietary MP exposure data, enabling a biomarker-anchored screening of dysbiosis-related signals associated with disease risk. Across the compiled disease cohorts, obesity provided the largest and most data-sufficient contrast in this dataset and was therefore selected as the anchor condition for threshold derivation. Receiver operating characteristic (ROC) analysis determined a model-derived screening cutoff for the F/B ratio (cutoff = 1.08), which was further used to define a three-tier risk classification system. A dataset comprising 558 dietary MP records from 129 studies across 25 countries was incorporated to predict food- and region-specific risk levels. Results reveal that seafood represents the dominant global dietary MP risk source, with mollusca contributing the highest predicted F/B-related risk signal. Spatial stratification indicates that the United States and Ecuador were classified into the highest model-derived tier based on national mean predicted F/B values, followed by high-risk regions such as Vietnam, Turkey, and Ireland, while countries in the Middle East and North Africa remain at moderate risk. These geographic contrasts suggest the combined influence of dietary structure and regional exposure intensity. Overall, we present a biomarker-anchored, machine-learning-enabled screening framework that links dietary MP exposure to predicted changes in a dysbiosis-associated indicator (F/B). This approach can support monitoring prioritization, identification of data gaps, and hypothesis generation for future validation studies.
15 June 2026
Read appraisal →Journal of hazardous materials
Prenatal 6:2 Cl-PFESA exposure and birth weight: A systematic review and meta-analysis
Most existing systematic reviews and meta-analyses on PFAS have focused on legacy compounds, yet with less attention on PFAS alternatives and mixtures. Chlorinated polyfluorinated ether sulfonic acids (Cl-PFESAs) have been introduced as PFOS alternatives. Among these, 6:2 Cl-PFESA has accumulated persistent environmental residues and has recently emerged as a significant health concern. We aim to evaluate the association between prenatal 6:2 Cl-PFESA exposure and birth weight using a meta-analysis. A systematic review identified 12 studies reporting birth weight in relation to 6:2 Cl-PFESA biomarker measures, including maternal serum and umbilical cord samples. Risk of bias and study sensitivity were assessed, and overall effects were estimated by random-effects model. Stratified analyses were conducted by study confidence, sample size, and exposure level. The overall pooled estimate showed no statistically significant association between 6:2 Cl-PFESA exposure and birth weight. However, subgroup analyses revealed a consistent reduction in birth weight per ln-unit increase in 6:2 Cl-PFESA in studies with large sample sizes (β = -20.05 g; 95% CI: -33.13, -6.97), high methodological confidence (β = -16.74 g; 95% CI: -26.42, -7.06), or higher exposure levels (β = -16.15 g; 95% CI: -25.43, -6.87), all showing low between-study heterogeneity. Sensitivity analysis excluding the study contributing most to heterogeneity indicated a 15.57 g decrease in birth weight per ln ng/mL increase in 6:2 Cl-PFESA (95% CI: -24.74, -6.40). Although current evidence suggests that prenatal 6:2 Cl-PFESA exposure may adversely affect fetal growth, well-designed prospective studies with standardized exposure assessments are needed to confirm this potential risk.
15 June 2026
Read appraisal →Journal of medicinal chemistry
Developing Predictive Models by Sharing Predictions - An Investigation of a Federated Learning Approach for ADMET Predictions
Machine learning models for ADMET prediction benefit from large, diverse data sets, yet such data are typically siloed across organizations. Federated learning (FL) enables collaborative modeling while preserving data privacy. Here, we investigate a student-teacher model (STM) framework in which organizations train internal models on proprietary data and share predictions on a public data set to generate pseudolabels for a centralized student model. As a proof of concept, 11 pharmaceutical companies contributed predictions for rat steady-state volume of distribution, yielding a pseudolabeled data set of ∼133,000 compounds. The resulting student model achieved performance comparable to individual teacher models on an external test set (RMSE ≈ 0.51 vs 0.47-0.61). Compared with FL approaches such as MELLODY and Effiris, STM offers a simpler workflow that avoids direct data sharing or iterative collaboration, providing a practical and scalable framework for secure cross-company model development.
13 June 2026
Read appraisal →The AAPS journal
New Frontiers of Drug Development Through the Use of New Approach Methodologies
On November 2025, AAPS PharmSci 360 convened a symposium that included experts in the application of various types of New Approach Methodologies (NAMs). They shared their experiences and insights on topics that included microphysiological systems (MPS), the use of 3D organoids, and in silico tools. MPS systems were explored from the perspective of their testing and qualification through an academia-industry-government tissue chip testing consortium whose mission is to perform context-of-use-based testing of MPS and conduct comparative evaluations to support the use of MPS systems in safety assessments. 3D organoids were described and insights shared on how and when they may provide an appropriate tool for evaluating drug safety and effectiveness, and their use in supporting personalized medicine. Regarding in silico tools, examples were provided to describe their growing utility in predicting drug toxicity, population variability, and its potential benefits over traditional evidence-based toxicology. These tools include in silico models (e.g., physiologically based pharmacokinetic models), Artificial Intelligence (AI), and Machine Learning (ML) which, unlike other aspects of model informed drug development, reduce reliance on predefined models and allows for the integration of diverse sources of data. Computationally, these tools can generate predictions in hours where it would otherwise have taken weeks or months (and extensive experimentation). Within this meeting report, we highlight the key issues discussed during that symposium and share additional aspects to consider when developing a NAMs-based roadmap for specific contexts of use.
1 June 2026
Read appraisal →Medicine
Environmental toxicant glyphosate induces cardiotoxicity: New insights from network toxicology, integrated machine learning, molecular modeling and multidimensional bioinformatics analysis
Glyphosate is one of the world's most widely used herbicide, yet the causal relationship between glyphosate and cardiotoxicity remains inadequate. To elucidate the cardiotoxic mechanisms of glyphosate, an integrative strategy was developed in this work, which combines computational toxicometry, predictive machine learning, and atomic-level molecular modeling. Initially, glyphosate's systematic toxicological profile was evaluated using ProTox-3.0 and ADMETlab 3.0. Candidate genes associated with glyphosate were obtained from 3 separate databases. Target genes associated with cardiotoxicity were integrated from 2 databases. Intersection analysis identified 113 core candidate genes associated with glyphosate and cardiotoxicity. Functional enrichment-related analyses were employed and revealed that the target genes were significantly involved in key biological processes such as ERBB3 signaling pathways. In order to refine the candidate genes to a more precise gene pool, net topology analyses were employed and identified 16 key feature genes. We subsequently integrated machine learning algorithms and validated the results using the GEO database, ultimately identifying 4 hub genes, AKT1, IL1B, BRCA1, and PTGS2. Finally, we applied molecular docking analysis and revealed strong binding affinities, which suggested direct interaction mechanisms in glyphosate-induced cardiotoxicity. In summary, our study establishes a comprehensive mechanistic framework for glyphosate-induced cardiotoxicity.
31 May 2026
Read appraisal →Journal of breath research
Investigating short-term lung inflammation using exhaled breath VOCs from exposure to candles: a randomized controlled crossover study among young mild asthmatics
Lung inflammation is associated with a response to pollution or irritants and is reported to release volatile organic compounds (VOCs) in breath, that can be monitored non-invasively. An approach to studying short-term lung inflammation is exposure to environmental emissions. This study used brief exposure to PM₂.₅ emissions from two types of burning candles to evaluate potential differences in airway inflammation among mildly asthmatic individuals. The aim was to explore VOCs linked to lung inflammation by exposure to candle emissions; emissions from candle A, a newly developed candle with a smaller emission to the surroundings and candle B, a well-characterised commercially available candle. Seventeen non-smoking asthmatics (11 female) with a mean age of 21.9 years participated in a randomised controlled double-blind crossover study including two exposure sessions: i) air with candle A emissions (mean PM2.554.1µg m-3), and ii) air with candle B emissions (mean PM2.598.4µg m-3). Participants underwent two five-hour, double-blind exposure sessions in a controlled climate chamber, each involving emissions from one of two candle types. Exhaled breath was collected at baseline (0 h), immediately post-exposure (5 h), and the following morning (24 h). VOCs were analysed using both a targeted panel of 23 predefined compounds and an untargeted workflow, which yielded 21 breath-derived VOCs after ambient-blank filtering. Statistical comparisons used Wilcoxon signed-rank tests with Benjamini-Hochberg correction. Across both exposure sessions, ten exhaled VOCs showed significant changes between baseline and immediate post-exposure (5 h), including increases in several aldehydes and decreases in selected sulfur-containing compounds. These changes were not present at 24 h, where samples resembled baseline profiles. The ten VOCs were identified from a combined dataset of 23 targeted compounds and 21 untargeted breath-derived VOCs; only the untargeted analysis yielded significant findings after correction. PCA demonstrated clear separation between baseline and 5 h samples, particularly for candle B, for which 0 h, 5 h and 24 h samples formed distinct clusters. This candle-specific response produced a multi-VOC signature capable of distinguishing individuals immediately after exposure, suggesting that exhaled VOCs can reflect short-term airway inflammatory responses. The findings suggest that an emission-specific, exhaled breath, multi-VOC profile can detect individuals exposed to inflammation triggers, such as candles. This indicates that VOCs could serve as biomarkers for detecting short-term airway inflammation, as evidenced by the systematic difference in VOCs observed 5 h after exposure to candle B. This research contributes to the advancing field of VOC-based health monitoring and might affect public health implications.
30 May 2026
Read appraisal →Analytical chemistry
Human-to-Rat Validation of PMI Biomarkers: A Bidirectional Cross-Species Metabolomics Study Reveals Asymmetric Translation
Accurately estimating the post-mortem interval (PMI, also known as time since death) remains a major challenge in forensic science due to substantial biological and environmental variability in human cases. While animal models provide controlled conditions for identifying PMI-related metabolic changes, their translational value for human forensics is unclear. This study bridges controlled rat metabolomics with real-world human autopsy samples using LC-MS-based metabolomics and machine learning. We analyzed femoral blood from 435 medico-legal autopsies (PMI 1-10 days) and employed two complementary modeling strategies. An untargeted Lasso model using all 3450 detected features achieved R2 = 0.51 and identified 51 consistently selected PMI-informative molecular features. To assess cross-species robustness, we reanalysed rat samples using the same analytical platform. Eighteen of the 51 human-derived features were detected in rats, showing clear PMI-dependent trends and strong correlations with PMI. Two metabolites achieved level 2 annotation and matched known PMI biomarkers from previous animal studies, demonstrating meaningful human-to-rat biochemical overlap. In contrast, a targeted Lasso model based on biomarkers identified in rats performed poorly in humans (R2 = 0.10), underscoring limited direct rat-to-human translation. Overall, human-derived biomarkers validated robustly in the controlled rat model proved far more reliable. These findings highlight both the potential and limitations of cross-species PMI biomarker discovery and emphasize the need for improved metadata, especially temperature history, to advance PMI modeling in forensic practice.
27 May 2026
Read appraisal →Journal of computer-aided molecular design
ADMET-vault: an interactive framework for real-time ADMET prediction and molecular optimization
Drug discovery pipelines face persistent challenges, as candidates that show promising results in early laboratory studies are found ineffective in clinical trials-often because their absorption, distribution, metabolism, excretion, and toxicity (ADMET) profiles were not well understood early enough. In this study, an interactive machine-learning based platform, ADMET-Vault, is presented, which brings ADMET prediction directly into the molecular design process, allowing researchers to assess and optimize drug candidates during the molecule design phase itself. ADMET-Vault integrates multiple molecular representations including physicochemical descriptors, molecular fingerprints, and graph neural network embeddings. The platform evaluated 12 benchmark datasets from Therapeutics Data Commons using scaffold-based validation. ADMET-Vault demonstrated consistent predictive performance across multiple ADMET endpoints. Strong results were observed for clearance related properties, intestinal absorption, hepatotoxicity, and aqueous solubility. Moreover, regression showed mean absolute error value that ranged from 0.28 (Caco-2 permeability, expressed as log cm/s) to 7.68 (plasma protein binding rate, expressed as % bound), with Spearman's rank correlation coefficients between 0.44 and 0.66. Classification tasks showed area under the receiver operating characteristic curve values ranged between 0.87 and 0.99. ADMET-Vault's main contribution is not just in its predictive accuracy. It combines reliable machine learning models with a live molecular editor. This allows researchers to see how a structural change impacts drug-like properties instantly. This interactive design-predict-optimize loop fills a long-standing gap in early-stage drug discovery workflows.
27 May 2026
Read appraisal →Renal failure
Development and validation of an early prediction model for bee-sting-induced acute kidney injury using machine learning
Acute kidney injury (AKI) is a severe and frequent complication following bee stings, with a reported incidence of 30-50%. Early identification is clinically challenging due to the delayed rise in serum creatinine. This retrospective study aimed to develop and validate an interpretable machine learning model for the early prediction of AKI in bee-sting patients. We included 305 patients admitted between January 2018 and September 2025, of whom 99 (32.5%) developed AKI according to Kidney Disease: Improving Global Outcomes (KDIGO) criteria. Least Absolute Shrinkage and Selection Operator (LASSO) regression was employed for feature selection from 33 baseline variables, identifying eight key predictors, including lactate, creatinine, and dark urine. Six machine learning algorithms - logistic regression, random forest, support vector machine, neural network, extreme gradient boosting (XGBoost), and light gradient boosting machine - were developed and evaluated. The XGBoost model demonstrated superior performance, achieving an area under the receiver operating characteristic curve of 0.964 and an accuracy of 92.6% on the test set. SHapley Additive exPlanations analysis was applied to interpret the optimal model, highlighting creatinine, lactate dehydrogenase, lactate, and dark urine as the most influential predictors. A clinically applicable nomogram was constructed and showed good calibration and utility. In conclusion, this study successfully developed a high-performance and interpretable XGBoost model for early prediction of bee-sting-induced AKI. The validated nomogram provides an accessible bedside decision-support tool to facilitate timely risk stratification and intervention, potentially improving patient outcomes.
26 May 2026
Read appraisal →Chemico-biological interactions
Carbon quantum dots exacerbate Parkinson's disease-like pathology through microglia-mediated neuroinflammatory responses with toxicological implications
The global burden of Parkinson's disease (PD) is projected to double by 2050, with early-onset cases demonstrating accelerated progression and limited therapeutic options. Recently, carbon quantum dots (CQDs) emerge as promising nanoplatforms for PD theragnostic, but their paradoxical role in exacerbating neuroinflammation and neurodegeneration shows the urgent need for mechanistic safety evaluations to bridge their therapeutic potential and neurotoxic risks. This study conducted a neuro-safety assessment of CQDs based on MPP+-induced mice and cell PD models to stimulate the theragnostic scenarios of CQDs in PD patients in the real-world. The results showed that CQDs not only promoted the onset and progression of PD-like damages in mice and cells but also exacerbated PD-related neurological dysfunction and structural damages in PD models. The microglia-mediated neuroinflammatory reactions were identified to be pivotal mediators of CQDs exacerbating PD-like damages based on machine learning (ML) approaches. According to these key toxic targets, we established molecular docking technology and network pharmacology to propose interventional strategies, including functional modification of CQDs and anti-neuroinflammatory drugs to alleviate the exacerbation of PD-like damages induced by CQDs, thereby ensuring the safe application of CQDs in the diagnosis and treatment of PD.
25 May 2026
Read appraisal →Analytica chimica acta
Optimized HSSE-GC-MS and machine learning for forensic gender classification of human scent from clothing
BACKGROUND: Clothes are considered a valuable piece of evidence in forensic investigations because of the underlying information that can be extracted from them. Human scent is one kind of trace evidence that can be extracted from clothes, which can link a specific person or location to the crime. This study optimized a headspace sorptive extraction (HSSE) method coupled with gas chromatography-mass spectrometry in combination with machine learning for objective discrimination of gender from human scent traces on clothing for forensic applications. RESULTS: The HSSE was optimized using the Box-Behnken design and Response Surface Methodology to maximize the Euclidean distance between the total ion sum spectrum (TIS) of two independent samples. The optimized extraction parameters were determined at 100 °C extraction temperature, 180 min extraction time, and 5 × 5 cm cloth dimension at m/z range 101 - 250 which produced an R2 of 0.9874. Method validation demonstrated an acceptable absolute error of 6.60%. Intra- and inter-day precisions were calculated for each m/z from 101 to 250 to remove outliers from the dataset for the multivariate analysis. Using the optimized extraction method, 62 samples were analyzed and the pre-processed TIS from the samples were subjected to various supervised machine learning algorithms. Support vector machine (SVM) and partial least squares-discriminant analysis (PLS-DA) models achieved perfect 100% accuracy for gender classification in both training and test sets. Random forest (RF) and linear discriminant analysis (LDA) also showed high test set accuracies of 93% and 86%, respectively. SIGNIFICANCE: This integrated approach, combining optimized stir bar HSSE-GC-MS with different machine learning algorithms using the TIS untargeted approach, provides a fast and automatic gender classification from human scent traces on clothes that can be used as a rapid screening technique in forensic investigations to narrow down suspect or victim profiles.
22 May 2026
Read appraisal →Journal of chemical information and modeling
In Silico Prediction of Potential Bile Salt Export Pump Inhibitors Using Pharmacophore Models and Consensus Machine Learning
The ATP-dependent bile salt export pump (BSEP) is a transporter responsible for moving bile salts from hepatocytes into bile canaliculi. Inhibition of BSEP is a known risk factor of cholestatic-drug-induced liver injury (DILI) probably caused by the accumulation of toxic bile salts in the liver. Since DILI is a major cause for attrition during drug development and postmarketing withdrawal or black box warnings, an IC50 of BSEP higher than 25 μM is advised. Here, we describe the investigation of several in silico approaches to map potential BSEP inhibitors. A consensus machine learning classification model is presented, which accurately predicts BSEP inhibitors and can assist in the prioritization of in vitro testing. An applicability domain was derived for the model, and it was validated by in vitro measurements of predicted BSEP inhibitors where the model was able to correctly flag eight out of 11 compounds.
13 May 2026
Read appraisal →Journal of ethnopharmacology
Senecio scandens Buch. - Ham.: A comprehensive review of botany, phytochemistry, biological activity, toxicity, quality control, application, and practical domain.
ETHNOPHARMACOLOGICAL RELEVANCE: Senecio scandens Buch.-Ham., a medicinal herb from the Asteraceae family, contains flavonoids, terpenoids, and alkaloids as its primary bioactive compounds. It is commonly used in the treatment of ocular diseases and dermatological conditions. However, with its expanding applications across various fields, growing attention has been directed toward the potential safety concerns associated with its use. AIM OF THE REVIEW: This article systematically reviews existing literature on S. scandens, examining its botany, phytochemistry, biological activity, toxicity, applications, and development utilization, to provide a theoretical basis for in-depth exploration of its pharmacological mechanisms and resource development. MATERIALS AND METHODS: This study conducted a systematic review of literature on S. scandens published from 1972 to 2025, based on databases including PubMed and Web of Science. Plants names are provided in accordance with "The Plant List" (https://wfoplantlist.org/). RESULT: Over 200 chemical constituents, primarily flavonoids, terpenoids, and alkaloids, have been isolated and identified from S. scandens, endowing it with diverse biological activities such as anti-inflammatory, antibacterial, antioxidant, and anti-tumor effects. To address potential hepatorenal toxicity from its pyrrolizidine alkaloids, relevant safety guidelines have been gradually established alongside various clinical dosage forms. It is also widely used in fields including medicine, animal husbandry, agriculture, and cosmetics. CONCLUSION: By systematically reviewing the botany, phytochemistry, biological activity, toxicity, application, and development and utilization of S. scandens, this paper identifies key utilization obstacles, proposes future research directions, and provides a theoretical framework for its further development.
12 May 2026
Read appraisal →Environmental science. Processes & impacts
Machine learning-driven QSAR models for the prediction of metabolic mechanisms and thyroid hormone-disrupting effects of emerging pollutants in the human body: a case study of bisphenol analogues
Research on theoretical prediction methods for elucidating the reaction mechanisms and thyroid hormone-disrupting effects of emerging pollutants in the human body faces significant challenges. The application of in silico methods utilizing machine learning is increasingly recognized as an effective strategy to determine the mechanisms of reactions catalyzed by human cytochrome P450 enzymes (CYPs) and to predict the adverse effects of emerging pollutants. Bisphenol analogues (BPs) represent one of the most important endocrine disruptors. These compounds can cause serious effects on the ecological environment and human health. Herein, density functional theory (DFT) calculations was employed to investigate the reaction mechanisms of the O-addition and H-abstraction of 20 BPs by human CYPs. A machine-learning-integrated quantitative structure-activity relationship (ML-QSAR) framework was established to predict the energy barriers of three types of reactions. Moreover, binding affinities were calculated between BPs (including their metabolites) and TRβ-LBD to evaluate their thyroid hormone-disrupting effects. The binding affinity data of BPs to TRβ-LBD were used as a training set to develop a ML-QSAR model for efficient prediction of potentially hazardous BPs. A ML-QSAR model with an R2 of 0.869 was established to predict their thyroid hormone-disrupting effects. The study enhances our understanding of CYP catalytic mechanisms involving BPs and provides a reference for the future design of environmentally friendly BPs. The predictive model developed for thyroid hormone-disrupting effects can aid in identifying the adverse effects of emerging BPs, supporting more effective management of toxic compounds.
4 May 2026
Read appraisal →Environmental science. Processes & impacts
Adverse outcome pathway-informed machine learning for predicting vascular toxicity of emerging organic pollutants
Emerging organic pollutants (EOPs) are increasingly detected in wastewater and pose potential vascular toxicity risks that remain inadequately assessed in current regulatory frameworks. This study developed an adverse outcome pathway (AOP)-informed machine learning approach to evaluate vascular toxicity for 312 EOPs. By integrating ToxCast high-throughput bioassay data with Morgan fingerprints, we trained fourteen multilayer perceptron (MLP) models targeting key events in AOP 509, including Nrf2 inhibition, oxidative stress, mitochondrial dysfunction, apoptosis, endothelial impairment, and angiogenesis disruption. Our optimized models achieved high predictive accuracy (70-95%), enabling activity classification for both tested and untested chemicals. The predicted activation profiles prioritized chemicals such as ketoconazole, sertraline, and miconazole, with literature evidence supporting their vascular toxicity potential. This AOP-guided modelling framework demonstrates how integrating mechanistic pathways with machine learning can inform chemical risk assessment and prioritization, supporting hazard evaluation and environmental decision-making for pollutants with limited toxicological data.
4 May 2026
Read appraisal →Annals of biomedical engineering
Stacking Ensemble Machine Learning for Cardiac Safety Assessment Using hiPSC-CM MEA Data
PURPOSE: Prediction of Torsades de Pointes (TdP) risk using hiPSC-CM assays remains challenging, as many models fail to capture nonlinear patterns and exhibit unstable performance across different drugs. We examined whether a stacking ensemble can improve robustness when only two simple Δ F P D c -derived predictors are available. METHODS: Two electrophysiological predictors were derived from MEA-based FPDc measurements: the maximum change ( Δ F P D c ) and the interpolated Δ F P D c at Cmax. These features were used to train a stacking model comprising random forest (RF), XGB, and a shallow artificial neural network (ANN). Performance was assessed on 16 unseen CiPA reference compounds using AUC, likelihood ratios, pairwise accuracy, and classification error. Class weighting addressed class imbalance, and likelihood-based metrics assessed diagnostic consistency. RESULTS: The stacking ensemble consistently outperformed all single classifiers. The XGB-based meta-classifier achieved perfect discrimination for both AUC1 and AUC2 (1.000), with pairwise accuracy reaching 1.000 and classification error remaining below 0.125 across repeated evaluations. CONCLUSIONS: Combining a simple MEA-derived feature set with a stacking architecture provides a more reliable framework for early TdP risk assessment than single-model approaches. External validation using additional MEA datasets and the integration of interpretable modeling strategies will be important for future translational use.
4 May 2026
Read appraisal →Journal of chemical information and modeling
Development, Evaluation and Application of a Multi-Representation Fusion Model for Accurate Prediction of Per- and Polyfluoroalkyl Substance (PFAS) Binding to Plasma Proteins
Per- and polyfluoroalkyl substances (PFAS) constitute a large and structurally diverse class of man-made chemicals. Their strong carbon-fluorine (C-F) bonds confer high environmental persistence, bioaccumulation, and various associated toxicities. As amphiphilic compounds, most PFAS bind to proteins and accumulate in protein-rich tissues, with such bioaccumulation exerting significant adverse impacts on human health. Accurate evaluation of the binding status between PFAS and proteins constitutes an essential step in health risk assessment. Traditional experiments and certain modeling approaches for analyzing PFAS bioaccumulation suffer from drawbacks such as time-consuming processes, high costs, or inadequate capture of molecular structural information, while existing machine learning-based prediction methods rely on single molecular representation, making it difficult to comprehensively encode the structural information on PFAS. Here, we propose MURNet, a multirepresentation fusion network model integrating chemical descriptors, 2D molecular graphs, and molecular fingerprints to predict PFAS-plasma protein binding. Compared with the state-of-the-art baseline models, MURNet achieves the optimal comprehensive performance. The multirepresentation fusion strategy generates higher-quality molecular features. Tanimoto similarity applicability domain analysis demonstrates MURNet's capability to reliably predict PFAS homologues. Case studies reveal the effectiveness of MURNet in screening PFAS with potential binding affinity to human serum albumin (HSA).
2 May 2026
Read appraisal →Nicotine & tobacco research : official journal of the Society for Research on Nicotine and Tobacco
Health Hazards of E-Cigarettes and Heated Tobacco Products: a Comprehensive Analysis of Hazardous Substances and Regulatory Gaps
INTRODUCTION: Ingredients and products of electronic cigarettes (ECs) and heated tobacco products (HTPs) formed during heating are aerosolized and inhaled. Hazard information on ingredients and emissions is often unavailable. Our objective was to identify and classify substances in the ingredients and emissions of ECs and HTPs based on scientific hazard classifications. METHODS: Ingredient data were collected from the European Union (EU) Common Entry Gate, a mandatory notification system for tobacco and related products. Substances in ECs and HTPs emissions were identified with a systematic literature review and classified into three categories based on existing health hazard classifications. Category 1 substances (i.e., highest potential health hazard) were regarded as high-priority substances for research and quantification. Categories 2 and 3 included potentially harmful substances and substances lacking sufficient toxicological data respectively. RESULTS: A total of 134 of the 1740 identified substances were classified in Category 1. Several substances possess Carcinogenic, Mutagenic, or Reprotoxic properties, prohibited as tobacco additives. Most substances lacked sufficient toxicity data and were classified in Category 3. CONCLUSION: Our classification can be used for monitoring and enforcement, ensuring proactive management of potential risks. Therefore, we recommend strengthening the enforcement of existing bans. Implications: This study identifies and classifies 1740 substances found in the ingredients and emissions of ECs and HTPs, including 134 hazardous chemicals, among which are carcinogens, mutagens, and reproduction toxicants. This indicates regulatory gaps and the need for stricter enforcement of existing bans. The findings provide a valuable tool for policymakers and public health professionals for monitoring and mitigating risks.
24 Apr 2026
Read appraisal →Toxicology mechanisms and methods
Introducing AutoML framework for drug-drug interaction prediction: application of AutoGluon.
Drug-drug interactions (DDIs) remain a major concern in medication safety. Although advanced artificial intelligence methods such as deep learning have improved DDI prediction, their adoption is limited by the need for specialized expertise and complex model development. This study introduces the first application of the AutoGluon AutoML framework to DDI prediction using molecular features, aiming to automate and simplify model development. A curated subset of 100,000 drug pairs from DrugBank was used, employing three molecular representations: 2D molecular descriptors, 2048-bit Morgan fingerprints, and their combination. Models were trained using AutoGluon-Tabular with no manual hyperparameter tuning. The descriptor-only model achieved the best performance, with 84.4% test accuracy and an AUC of 0.916, outperforming fingerprint-based and hybrid models. Feature importance analysis identified key physicochemical and topological descriptors-such as drug-likeness, electrotopological indices, and hydrophobic surface area-as critical predictors of DDIs. These results demonstrate that AutoML can extract chemically meaningful patterns while reducing technical barriers. Overall, our results validate AutoGluon as a scalable approach to DDI prediction that provides chemically meaningful, feature-level interpretability, and lay the groundwork for future applications involving larger datasets and more complex chemical representations.
15 Apr 2026
Read appraisal →Tobacco control
Nitrosamine and nicotine exposure after switching from filtered to unfiltered cigarette smoking: a cross-over clinical trial
BACKGROUND: The cellulose acetate filter is a plastic attachment on nearly all commercial cigarettes sold worldwide. It is the main component of discarded cigarette butts, and the most prevalent waste item collected during urban and beach cleanups. This waste leaches toxic chemicals, including nicotine, metals and tobacco-specific nitrosamines and contributes to environmental microplastic pollution. There is growing international interest in reducing plastic waste from single-use, non-essential products such as cellulose acetate cigarette filters. Public health and environmental advocates recommend a ban on the sale of filtered cigarettes to reduce environmental pollution caused by discarded cigarette butts and to discourage cigarette smoking. RESEARCH QUESTION: What potential health and behavioural implications might arise among people who smoke if filtered cigarettes are removed from the market. METHODS: We conducted a cross-over, randomised clinical trial of 29 people who smoke to evaluate changes in biomarkers for nicotine and tobacco-specific nitrosamine exposure when switching between filtered and unfiltered cigarette smoking. RESULTS: Although unfiltered smoking showed a higher trend in the geometric means of 4-[(methylnitrosamino])-1-[3-pyridyl]-1-butanol (NNAL) compared with filtered cigarettes, regression models for urinary cotinine and NNAL showed no significant differences when switching between filtered and unfiltered cigarettes. CONCLUSION: This proof-of-principle study suggests there is no increased risk to people who smoke when switching to unfiltered cigarette smoking. Although larger studies might provide more evidence regarding unfiltered cigarette smoking, banning the sale of filtered cigarettes may be an important policy intervention to both reduce hazardous tobacco waste and discourage smoking.
15 Apr 2026
Read appraisal →Journal of chemical information and modeling
ToxPLTC: Peptide Toxicity Prediction by Integrating Pretrained T5 Protein Language Model and Text Convolutional Neural Network
Peptide-based therapeutics show promising potential in treating a range of diseases, such as diabetes, cancer, and chronic pain. However, critical challenges, including peptide toxicity, immunogenicity, and stability deficiencies of peptides, have become major obstacles to their direct clinical application. Traditional toxicity testing methods based on a wet lab are not only costly but also time-consuming. In contrast, the classification model based on deep learning provides a new technical path for the efficient identification of peptide toxicity. In this study, we propose a deep learning framework called ToxPLTC, which employs the ProtT5 protein language model based on the Transformer architecture for pretraining peptide sequences, adopts the borderline SMOTE algorithm to handle an imbalanced training set data, and utilizes a text convolutional neural network combined with a fully connected layer for classification. Additionally, visualization analysis, motif analysis, and mutation-scan analysis are performed to understand the function of each module and enhance the interpretability of our model. The applicability domain is constructed based on the K-NN strategy to define the effective prediction range of our model to ensure the reliability of model predictions. The ToxPLTC model achieves a balanced accuracy of 93.02% on independent test set 1 and 88.04% on independent test set 2. Experimental results demonstrate that our model outperforms existing models on independent test sets and has good generalization ability. The ToxPLTC model possesses great potential as a valuable and robust tool for peptide-based drug development. The source data sets and codes can be available at the following GitHub repository: https://github.com/yunyunliang88/ToxPLTC.
14 Apr 2026
Read appraisal →Journal of chemical information and modeling
EviCYP: In Silico Prediction of Cytochrome P450 Substrates Based on Vector Quantization and Evidential Deep Learning
The accurate identification of cytochrome P450 (CYP) substrates is crucial in drug discovery and safety assessment, as these enzymes mediate the metabolism of most clinical drugs. However, existing computational models are often limited by data quality issues and lack the ability to quantify prediction uncertainty, hindering their reliable application. To address these challenges, we present EviCYP, a novel prediction framework that integrates evidential deep learning with vector quantization (VQ). We first constructed a high-quality data set by curating 4388 substrates and 2880 nonsubstrates from 1629 publications, and supplemented it with 3728 pseudonegative samples, resulting in 10,996 samples spanning nine major CYP isoforms. The EviCYP architecture processes multimodal molecular representations and enzyme sequences through dedicated encoders, compresses features via VQ to reduce redundancy, and employs an evidential layer to output both class probabilities and an uncertainty estimate. On an internal test set, EviCYP achieved an average AUROC of 0.9500. Notably, the model's uncertainty quantification is highly reliable, with high-uncertainty predictions strongly correlating with classification errors. This work provides a robust and trustworthy computational tool for CYP substrate prediction.
14 Apr 2026
Read appraisal →Journal of chromatography. A
Machine learning-assisted HS-GC-IMS for discrimination and traceability of baby bottles based on volatile fingerprints
Ensuring the safety of food contact materials, particularly baby bottles, is crucial for infant health. In this study, a comprehensive analytical strategy integrating headspace gas chromatography-ion mobility spectrometry (HS-GC-IMS) with machine learning-based chemometrics was developed to profile volatile organic compounds (VOCs) and achieve discrimination and traceability of materials and brands among nine baby bottles. A total of 87 VOCs were identified, with aldehydes, esters, alcohols, ketones being the most diverse groups, and aldehydes exhibiting the highest concentrations. Visual representations were established, clearly illustrating the fingerprint profiles of the various baby bottle samples. Unsupervised models, principal component analysis (PCA) and hierarchical cluster analysis (HCA), revealed distinct clustering patterns. In contrast, supervised models, including Orthogonal Partial Least Squares-Discriminant Analysis (OPLS-DA), Decision Tree (DT), Random Forest (RF), and Support Vector Machine (SVM), and deep learning models such as Convolutional Neural Network (CNN) and Transformers, were constructed to predict brand classification. Among them, RF was selected as the optimal model due to its superior performance-utility balance. Key discriminant markers were further determined by integrating the Variable Importance in Projection (VIP) scores from OPLS-DA with the feature importance from the RF model, significantly improving classification efficiency. This research proposes an effective and reliable analytical framework for rapid identification and traceability of baby bottles, offering valuable insights for anti-counterfeiting and safety assurance of food-contact materials.
12 Apr 2026
Read appraisal →Environmental science & technology
Molecular Representation and Closed-Loop Validation for Toxicity Assessment of Organic Compounds in Ambient Air PM2.5
Although the health impacts of fine particulate matter (PM2.5) are primarily attributed to its chemically diverse composition rather than to mass concentration, assessing the toxicity of PM2.5 constituent compounds remains highly challenging due to chemical complexity and limited experimental scalability. This study introduces an interpretable machine learning (ML) framework using a curated A549 cytotoxicity data set (19,841 compounds) that integrates customized Molecular Access System (MACCS) fingerprints, six base models, and a stacking ensemble meta-model, all optimized with a biobjective strategy to assess the toxicity of organic compounds in PM2.5. The stacking ensemble model demonstrated satisfactory performance (test AUC > 0.8), exhibiting good generalization, adaptability, and robustness. Nontargeted analysis generated and experimentally validated a prediction set from the Hong Kong PM2.5 samples (51 of 387 compounds confirmed from 13 classes), demonstrating broad applicability on an independent Nanjing PM2.5 set (572 compounds). Key substructures driving toxicity, identified through a SHapley Additive exPlanations (SHAP) analysis and cell experimental validation, revealed that PM2.5 compounds with aromatic rings and nitrogen-based functional groups (e.g., aromatic α,β-unsaturated ketones, aromatic amines, aromatic nitro compounds, and tertiary amines) likely contribute to high toxicity. The derived "structure-toxicity" rules narrow the search space from thousands of compounds to those containing critical substructures, enabling efficient prioritization of toxic components and providing a foundation for improving model specificity and predictive accuracy in future studies.
8 Apr 2026
Read appraisal →Journal of ethnopharmacology
Unveiling the genus Taraxacum: From folk medicine to chemodiversity-driven pharmacological and toxicological outcomes-A systematic review
ETHNOPHARMACOLOGICAL RELEVANCE: The genus Taraxacum (dandelion), belonging to the Asteraceae family, encompasses over 2800 species predominantly found in temperate to subtropical regions of the Northern Hemisphere. Dandelion has a long history as a functional food and traditional medicine in numerous countries, including China, Turkey, Spain, Italy, etc. Traditional application records indicate that dandelion might be used extensively to treat kidney stones, constipation, eczema, ulcers, inflammation, etc. However, the therapeutic application of different Taraxacum species varies significantly across regions, which may be influenced by local cultural and medical traditions. AIM OF THE REVIEW: This review provides a comprehensive summary of recent advancements in traditional applications, phytochemistry (highlighting 184 newly identified compounds), pharmacology, and toxicology, offering novel insights into both genus-wide characteristics of Taraxacum (covering 55 species) and species-specific differences. MATERIALS AND METHODS: A comprehensive search for relevant data was conducted using keywords such as "Taraxacum", "dandelion", "Taraxaci Herba", "pu gong ying", and others across multiple scientific databases, including Google Scholar, Web of Science, PubMed, ScienceDirect, Baidu Scholar, CNKI, online ethnobotanical databases, ethnobotanical monographs, The Plant List, Dictionary of Chinese Ethnic Medicine and considered data from 1972 to 2025. World Flora Online (http://www.worldfloraonline.org/) was utilized for confirming the correct species names of Taraxacum genus. The gathered information on the ethnobotanical usages, phytochemistry, pharmacology, and toxicology is comprehensively summarized in this paper. RESULTS: The literature review indicates that dandelion, a medicinal and edible plant, has been extensively utilized in traditional Chinese medicine since the 7th century. In Europe, its edible and medicinal properties began to be acknowledged and applied during the Middle Ages. In ethnomedical practice, dandelion has been used as diuretic, laxative, anti-inflammatory remedy, digestive aid, etc. Building on the 197 compounds reported from T. officinale, this review highlights over 180 newly identified compounds from 19 Taraxacum species, expanding the genus-wide catalog to more than 380 constituents. Recent toxicological and safety evaluations corroborate that T. officinale F.H.Wigg., T. mongolicum Hand.-Mazz., and T. lacistrum Sahlin possess a favorable safety profile. CONCLUSIONS: The genus Taraxacum is a promising medicinal and edible plant, exhibiting a range of biological activities, including anti-cancer, anti-inflammatory, and antioxidant effects. These properties align with its ethnobotanical usages, confirming its traditional use as diuretic, laxative, anti-inflammation, and digestive aid. However, research on its effects on neurodegenerative and cardiovascular diseases remains limited. Future studies should focus on in vivo experiments to explore the material basis and mechanisms underlying the therapeutic effects of dandelion species. Furthermore, current phytochemical and pharmacological research has overlooked several Taraxacum species with well-documented traditional uses. Investigating these understudied species is essential to fully elucidate the potential therapeutic value of the genus Taraxacum.
6 Apr 2026
Read appraisal →Cardiovascular toxicology
Association Between Environmental Chemicals Phthalates and Stroke Among the US Adults: A Cross-Sectional Study
Stroke has a multifactorial etiology, and phthalates, as widely used environmental chemicals, may play an underexplored role in cerebrovascular health. This study aimed to evaluate the association between urinary phthalate metabolites and stroke risk using data from the National Health and Nutrition Examination Survey (NHANES), complemented by network toxicology, molecular docking, and machine learning approaches. We analyzed data from 10,771 adult participants in NHANES (2003-2018) who had complete records of stroke status, urinary phthalates, and clinical covariates. Multivariable logistic regression and restricted cubic spline (RCS) models were used to assess associations. We constructed protein-protein interaction networks and performed molecular docking to identify potential toxicological targets. Five machine learning algorithms were employed to develop predictive models for phthalate-related stroke risk. Higher levels of MECPP, MCPP, MNBP, and MEOHP were significantly associated with increased stroke prevalence. Network toxicology identified SRC as a key target, and molecular docking confirmed strong binding affinity between MBZP and SRC. Among the machine learning methods, Random Forest achieved the highest predictive performance (AUC = 0.833). Notably, moderate coffee consumption (≤ 170 g/day) was significantly associated with a reduced risk of stroke (OR = 0.63, 95%CI: 0.51-0.77), suggesting potential dietary mitigation against phthalate-related stroke risk. Our findings demonstrate that certain phthalates may increase stroke risk via molecular pathways involving SRC. Furthermore, moderate coffee consumption may mitigate the adverse cerebrovascular effects of these environmental pollutants.
5 Apr 2026
Read appraisal →Environmental geochemistry and health
Deciphering stone mining-induced hazardous heavy metal contamination in agricultural soils using source attribution, health-dietary risk analysis, and machine learning-driven insights
Mining activities significantly contribute to the release of hazardous heavy metals (HHMs) into agricultural soils, especially where stone extraction is intensive. This study assessed metal content, source identification, and possible health risks associated with HHMs (Cd, Pb, Cr, Cu, Ni) in stone mines affected agricultural soils of India. The results revealed that Zone 1 exhibited higher concentrations of Cr (308.95 ± 23.03 mg/kg), Ni (117.11 ± 16.21 mg/kg), and Cd (4.21 ± 1.26 mg/kg), surpasses the threshold limit set by WHO/FAO for agricultural soils. In contrast, Zone 2 had considerably lower HHMs concentrations for Cr (60.24 ± 10.68 mg/kg), Ni (31.61 ± 7.52 mg/kg) and Cd (0.66 ± 0.13 mg/kg). Geospatial distribution and self-organizing map (SOM) indicated spatial clustering of HHMs, with elevated concentrations centered in the southern part of the study area. Positive matrix factorization showed that geogenic sources and mining activities affect significantly to HHMs pollution. Monte Carlo simulation revealed that children are more prone to exposure to HHMs than adults, with ingestion identified as the primary exposure route. Hazard quotient values calculated from free ion activity model exceeded the threshold limit for Pb (1.59E+00) and Cd (6.32E-01). Severity adjustment margin of exposure highlighted severe human health risk for Cr and Ni. Random forest analysis indicated that bioavailable HHMs significantly influence the accumulation of metals in rice tissues which elevating dietary risk. The integration of robust statistical tools offers insights into the dynamics of HHMs contamination and provide policymakers and environmental scientists with useful information to mitigate the potential risks caused by mining on nearby agricultural lands.
5 Apr 2026
Read appraisal →Journal of hazardous materials
Structuring large language models for chemical health risk reasoning in environmental and occupational exposure
Large language models (LLMs) have revolutionized numerous scientific fields. However, their application in environmental health remains limited, as they are prone to hallucinations and struggle with reasoning over complex, real-world exposure data. Chemical risks, especially from rare or poorly studied compounds, span diverse occupational and everyday scenarios, where structured knowledge is often incomplete and poorly integrated. To address these challenges, we present a health risk assistant that integrates deep learning-based compound property prediction, retrieval-augmented generation (RAG), and prompt engineering into a unified reasoning framework. Central to our design are two synergistic strategies: semantic structuring, which reorganizes heterogeneous knowledge into context-consistent and semantically rich profiles, and domain-aware prompting, which guides large language models through structured and role-specific reasoning aligned with toxicological goals. The framework enables the system to map physicochemical attributes to exposure pathways and disease outcomes with improved relevance and fidelity. Evaluated on a 100-question benchmark spanning chronic toxicity, occupational hazards, and exposure-specific risks, our LLM-RAG system consistently outperforms baseline models in correctness, faithfulness, and domain-specific helpfulness. Beyond answering queries, our approach demonstrates how integrated knowledge curation and tailored prompting can enable trustworthy and scalable AI reasoning for public and environmental health protection.
4 Apr 2026
Read appraisal →