Research Appraisals
Evidence-based critical appraisals of the latest medical research, systematically evaluated using Oxford CEBM methodology.
Showing 23 appraisals
Archives of toxicology
A review of machine learning in toxicology: current practices and reporting gaps
In recent years, machine learning and artificial intelligence approaches have been increasingly applied in the context of toxicological risk assessment. Many published overview, review, and comment papers discuss advantages, disadvantages, success stories, and open challenges for the application of machine learning models in toxicology. Machine learning methods using information from in vitro experiments can help to avoid animal experiments, thus allowing for larger numbers of experiments to be conducted. Drawbacks of machine learning models are the lack of mechanistic interpretability and the need for large amounts of high-quality data. In this work, we present a literature review of papers indexed in PubMed or published in the journal Computational Toxicology in the years 2022 to 2024, to assess the usage of machine learning methods in toxicology as well as the practices in reporting of methods and corresponding results. We do not address the suitability or the performance of methods, which is impossible to assess objectively without reanalysis on raw data, but focus on common practices and gaps in reporting. Major results are that many different machine learning methods are used in toxicology, often with appropriate internal validation. However, in only half of the cases, interpretation methods are used to address the problem that these models often make predictions as a black box. Moreover, there are very frequent gaps in reporting, in particular related to handling of missing values, and availability of data and code. Thus, this review can serve as a starting point for further tailored methodological research and guidance.
2 Aug 2026
Read appraisal →Journal of pharmacokinetics and pharmacodynamics
Diffusion models for virtual populations and pharmacometric simulations
To evaluate diffusion model-based artificial intelligence approaches for generating virtual populations with physiological determinants of drug dosing (PDODD) and pharmacokinetic (PK) profiles. A denoising diffusion probabilistic model (DDPM) was applied to a 31-variable dataset of PDODD covariates (18 continuous, 13 binary) from the National Health and Nutrition Examination Survey and compared to a tabular variational autoencoder (TVAE). For nivolumab PK data (12,000 patients, 13 time points, 5 covariates), sequence-based diffusion model (SDM) and a time-aware diffusion model (TDM) with temporal self-attention were evaluated. The predictive performance of the TDM was evaluated by imputing masked time points. All models were trained and tested on 80%:20% partitions of the data using univariate, bivariate, and multivariate distributional similarity metrics. The diffusion model satisfactorily approximated the univariate distributions of continuous PDODD biomarkers (mean Kolmogorov-Smirnov D-statistic, KSD = 0.014), disease status frequencies (mean absolute error, MAE = 0.31%), and preserved bivariate correlations (MAE = 0.033). DDPM outperformed TVAE for categorical variables (0.31% vs. 1.07% MAE) and correlation (0.033 vs. 0.091 MAE). For nivolumab PK, SDM has KSD of 0.047 and a relative error of 1.36%. TDM accurately imputed missing PK timepoints (KSD = 0.014), reconstructing masked Day 1, Peak concentration (Cmax) Dose-9, and Terminal phase concentrations with MAE of 0.43%, 0.25%, and 0.76%, and correlations ≥ 0.999. Diffusion models demonstrated strong performance in generating cross-sectional PK covariate data and longitudinal PK profiles, capturing complex distributional and temporal dependencies. Diffusion-based approaches provide a flexible and robust framework for virtual simulations in pharmacometrics.
1 Aug 2026
Read appraisal →Proceedings of the National Academy of Sciences of the United States of America
Legal infrastructure for transformative AI governance
Most of our AI governance efforts focus on substance: What rules do we want in place? What limits or checks do we want to impose on AI development and deployment? But a key role for law is not only to establish substantive rules but also to establish legal and regulatory infrastructure to generate and implement rules. The transformative nature of AI calls especially for attention to building legal and regulatory frameworks. In this Perspective, I review three examples: the creation of registration regimes for frontier models; the creation of registration and identification regimes for autonomous agents; and the design of regulatory markets to facilitate a role for private companies to innovate and deliver AI regulatory services.
29 July 2026
Read appraisal →PLoS medicine
How to benchmark medical AI agents
Medical artificial intelligence research is shifting from single-task models toward multimodal large language model-based agents for complex clinical workflows, requiring benchmarks that assess clinical reasoning, process safety, and resource stewardship rather than final outputs alone.
12 July 2026
Read appraisal →Journal of medicinal chemistry
Machine Learning-Based Models to Predict Drug-Induced Liver Injury (DILI) to Assist Medicinal Chemistry
Drug-induced liver injury (DILI) is a leading cause of drug failure and post-market withdrawals. Traditional preclinical methods fail to detect up to 40-45% of clinical hepatotoxicity cases. Computational approaches, particularly those based on machine learning and deep learning (DL), are emerging as promising tools to support medicinal chemistry and early drug discovery, though their predictive capabilities remain under active investigation. In this perspective, we review the development of DILI annotation data sets, tracing their growth from small collections to large, comprehensive resources. We also outline the evolution of computational methods, from simple descriptor-based models to advanced DL and ensemble approaches that incorporate interpretable features. Finally, we highlight recent efforts to integrate standardized causality frameworks, pharmacogenomics, and mechanistic models, aiming to connect computational advances with clinical relevance. This perspective provides valuable insight for researchers and promotes the development of more robust and consensual DILI prediction strategies.
10 July 2026
Read appraisal →Gut
Digital twins in IBD-bridging data, biology and trial innovation
Clinical trials in IBD face difficulties of escalating complexity, high costs and challenges in recruitment. Digital twins are virtual, data-driven replicas of individual patients that model disease trajectories and treatment responses, which offer a potential innovative change in the conduct of clinical trials in IBD. Built from multimodal datasets integrating clinical, molecular, imaging and real-world data, digital twins can generate synthetic control arms, enable adaptive randomisation and predict disease relapse or treatment response. Early studies across oncology, cardiology and endocrinology demonstrate their feasibility and potential to improve statistical power while reducing patient burden. However, the integration of digital twins into clinical trials in IBD will require rigorous validation frameworks, transparent data governance and attention to algorithmic bias and consent. In this review, we explore how digital twins may transform IBD research-from in silico simulation to adaptive, patient-centred trial design-and outline the regulatory, ethical and logistical challenges to be considered in order to successfully integrate them into future trials.
10 July 2026
Read appraisal →Molecular pharmaceutics
Computational Identification of Active Drug Metabolites for Human Protein Targets
Understanding drug metabolism helps mitigate toxicity risks and anticipate pharmacological effects beyond the parent compound. Here, we explore high-confidence in vitro human bioactivity annotations from the public domain associated with a curated data set of drug-metabolite pairs. Our findings show that 26% of all drug-metabolite-target combinations contain metabolites with retained or increased bioactivity relative to their parent drugs, showing a high proportion of active metabolites with potencies below 100 nM. Moreover, these active metabolites display similar predicted physicochemical and ADME properties to their parent drugs, indicating that their biological activity may be preserved in vivo. Two key features of the identified active metabolites are noted: (1) the metabolite is more likely to be formed via an O-demethylation or an N-demethylation at a tertiary nitrogen, and (2) these drug-metabolite pairs are frequently active against the membrane receptor target class. To support further research in this area, we make our data compilation, including pairs with identified active metabolites and recorded bioactivity data, openly available.
7 July 2026
Read appraisal →JMIR dermatology
The Rise in Homemade Sunscreen Trends and Future Impacts on Skin Cancer Risk: Systematic Review
BACKGROUND: The trend of homemade sunscreen recipes has rapidly gained popularity over the last decade, being largely fueled by social media influencers, natural health blogs, and the growing mistrust of large health organizations like the US Food and Drug Administration (FDA). Consumers are increasingly drawn to products labeled "natural," "organic," "vegan," and "cruelty-free," often conflating these terms with safety and effectiveness. However, when applied to sun protection, these assumptions can be dangerously misleading. According to the FDA, there is no verified mathematical formula that can be used to determine an accurate sun protection factor (SPF) rating for the amount of zinc oxide used in many of the recipes found online. OBJECTIVE: The objective of this review is to examine the rising popularity and efficacy of homemade sunscreens compared to commercial sunscreens and highlight the potential public health implications of skin cancer risk related to homemade sunscreen use. METHODS: Multiple databases were searched to find current and relevant literature analyzing the social media impact relating to homemade sunscreens and the efficacy of homemade sunscreens compared to commercially available products. Digital content that included or discussed homemade sunscreen recipes was included as well to provide examples of information trending online. RESULTS: One study analyzing social media trends found that many strong, unverified claims relating to homemade sunscreen versus commercial sunscreen use are being presented in the media, potentially influencing public knowledge. Additional studies analyzing the components of homemade sunscreen recipes compared to components in commercially available products found that homemade sunscreens offer little to no sun protection, therefore increasing the risk of UV damage for individuals using homemade products. CONCLUSIONS: Based on the presented evidence, the adoption of unregulated homemade sunscreens presents a substantial threat to public health. Misinformation spread through social media and influencer culture may unintentionally contribute to an increase in UV-related skin cancers. Greater efforts are needed from health care professionals, regulatory bodies, and online platforms to educate the public and promote the use of scientifically validated sun protection methods.
4 July 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 →Journal of controlled release : official journal of the Controlled Release Society
Toward animal-free approaches in subcutaneous drug delivery: A pain management case study.
Subcutaneous (s.c.) administration offers practical advantages for long-acting drug delivery, yet its complex tissue environment, lack of standardized models, and limited regulatory guidance pose challenges for predicting pharmacokinetics (PK). Here, we present a complementary set of experimental and computational approaches to reduce reliance on animal testing while maintaining translational relevance. Ex vivo skin models captured the influence of tissue variability, while 3D-printed constructs and molded agarose hydrogels enabled standardized, reproducible in vitro drug release studies. In parallel, machine learning models trained on curated rodent datasets predicted key PK parameters with strong agreement to in vivo data, providing an alternative that bypasses additional animal experiments and resource-intensive assays. Finally, a liposomal carprofen formulation case study demonstrated the translational potential of these methods in veterinary applications. Together, these strategies illustrate routes toward animal-free drug development: simplified experimental systems that mimic s.c. release and computational models that predict systemic PK. By combining them conceptually, we outline a framework that advances the 3Rs principles while supporting mechanistic and predictive understanding of s.c. drug delivery.
10 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 →European journal of heart failure
Surrogate markers of clinical endpoints in clinical trials of heart failure.
AIMS: We aimed to determine whether treatment related changes in surrogate markers predict longer-term therapeutic effects on all-cause mortality and heart failure (HF) hospitalization in randomized trials of HF. METHODS: Systematic literature search included randomized trials of interventions in patients with HF until October 2024. Random-effects meta-regression models with inverse variance weighting were calculated between fifteen surrogate markers and the clinical endpoints of (i) all-cause mortality and (ii) HF hospitalization. The degree of heterogeneity of each model explained by the surrogate marker was determined with R2. Surrogate threshold effects (STEs) were also calculated. RESULTS: Ninety-six randomized trials with median follow-up 12 months were included, enrolling a total of 120 304 patients with HF. Treatment related changes in natriuretic peptides (NPs) were weakly correlated with all-cause mortality (64 comparisons, P = .002, R2 = 12%, and STE = -34%) and HF hospitalization (41 comparisons, P < .001, R2 = 32%, and STE = -30%). Changes in LVEF were moderately correlated with mortality (69 comparisons, P < .001, R2 = 59%, and STE = +6%) and strongly correlated with HF hospitalization (45 comparisons, P < .001, R2 = 90%, and STE = +2%), although this finding was only observed in trials of HFrEF. LV end-diastolic diameter was correlated with HF hospitalization but not mortality. Other echocardiographic markers were not predictive of clinical endpoints. Treatment related changes in patient reported outcomes and exercise capacity were either weakly correlated with clinical outcomes or derived from small clinical trials. CONCLUSION: In HF trials, most surrogate markers, including NPs and echocardiographic measures were only weakly or moderately correlated with treatment effects on all-cause mortality. There is insufficient evidence to support the use of a single biomarker or echocardiographic measure in regulatory trials of HF.
14 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
Long Gu (Os Draconis): Textual research, modern scientific evaluation, and quality control challenges.
ETHNOPHARMACOLOGICAL RELEVANCE: Long Gu (Os Draconis), a mineral medicine with over 2000 years of use in traditional Chinese medicine, is facing a critical juncture. Its identity has been clarified scientifically-from a mythologized "dragon bone" to a fossilized mammalian bioapatite-yet its sustainable clinical application is threatened by resource depletion, widespread domestic and international adulteration, and the lack of modern quality standards. AIM OF THE STUDY: This review aims to systematically analyze the evolution of Long Gu through historical materia medica, define its authentic characteristics using modern science, diagnose the root causes of its current market and regulatory crisis, and propose a robust, multi-dimensional framework for quality evaluation and sustainable sourcing. MATERIALS AND METHODS: A systematic review of Chinese materia medica literature from the Pre-Qin to Qing Dynasties was conducted. Historical textual research was integrated with evidence from modern mineralogy, paleontology, geochemistry, and pharmacology to validate traditional knowledge and establish scientific identification criteria. RESULTS: The understanding of Long Gu evolved from mythological origins to a scientific conclusion in the Ming Dynasty, with Li Shizhen correctly identifying it as fossilized ancient mammalian bones. Modern research confirms authentic "Wuhua Long Gu" from the Shanxi-Shaanxi region is primarily carbonated hydroxyapatite, characterized by high porosity (>35%), significant hygroscopicity, and a unique trace element profile (Zn, Sr). These properties underpin its advertised therapeutic effects. However, the market is now saturated with dangerous adulterants, including fluoride-rich dinosaur fossils and processed modern bones, due to resource exhaustion and inadequate pharmacopoeial standards focused solely on calcium content. CONCLUSION: The future of Long Gu depends on transcending outdated quality controls. Ensuring its efficacy and safety necessitates a paradigm shift to a comprehensive standard system that combines paleontological and geochronological origin verification, quantitative microstructural analysis, chemical fingerprinting of key components, and strict safety monitoring. Complementary research into biomimetic synthesis and accelerated mineralization is urgently needed to develop sustainable alternatives for this invaluable yet endangered medicinal resource.
12 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 →Clinical pharmacology and therapeutics
Off-the-Shelf Large Language Models for Guiding Pharmacoepidemiological Study Design
This study aimed to assess the ability of two off-the-shelf large language models, ChatGPT and Gemini, to support the design of pharmacoepidemiological studies. We assessed 48 study protocols of pharmacoepidemiological studies published between 2018 and 2024, covering various study types, including disease epidemiology, drug utilization, safety, and effectiveness. The coherence (i.e., "Is the response coherent and well-formed, or is it difficult to understand?") and relevance (i.e., "Is the response relevant and informative, or is it lacking in substance?") of the large language models' responses were evaluated by human experts across seven key study design components. Coding accuracy was assessed. Both large language models demonstrated high coherence, with over 90% of study components rated as "Strongly agree" by experts for most categories. ChatGPT achieved the highest coherence for "Index date" (97.9%) and "Study design" (95.8%). Gemini excelled in "Study outcome" (93.9%) and "Study exposure" (95.9%). Relevance, however, was more variable, with ChatGPT aligning with expert recommendations in over 90% of cases for "Index date" and "Study design" but showing lower agreement for covariates (65%) and follow-up (70%). Coding agreement percentages reveal varying levels of concordance, with the Anatomical Therapeutic Chemical classification system coding system demonstrating the highest agreement at 50% with experts. In contrast, the Current Procedural Terminology and International Classification of Diseases systems showed agreements of 22.2% and 20%, respectively. While ChatGPT and Gemini show promise in certain tasks supporting pharmacoepidemiological study design, their limitations in relevance and coding accuracy highlight the need for critical oversight by domain experts.
4 May 2026
Read appraisal →Journal of chemical information and modeling
DWPL-GCNMF: Structure-Aware Dynamic Weighted Pseudo-Label Learning for Adverse Drug Reaction Prediction
Adverse drug reaction (ADR) prediction is typically formulated as drug-ADR association learning on extremely sparse, positive-unlabeled matrices, which limits generalization and makes naive pseudolabel expansion prone to confirmation bias. Here, we propose DWPL-GCNMF, a structure-aware semisupervised framework that combines graph convolutional network (GCN)-based structural embeddings learned from a drug-protein knowledge graph with matrix factorization scoring and introduces a prediction-consistency-driven dynamic weighted pseudolabel learning strategy to expand training pairs while down-weighting unreliable pseudolabels; multiple base models are further fused to improve prediction stability. We evaluate DWPL-GCNMF on a DrugBank-derived benchmark (1,177 drugs × 4,247 ADRs; ∼2.6% observed associations) and an external SIDER benchmark (1,080 drugs × 5,579 ADRs; ∼2.3% observed associations) to assess robustness under different data sets and increased sparsity. DWPL-GCNMF consistently outperforms representative baselines, achieving AUPR/F1 of 0.6656/0.6194 on DrugBank and 0.6106/0.5816 on SIDER, while also improving top-K prioritization (e.g., Recall@15 up to 0.8078 and 0.8146, respectively), indicating that structure-aware representation learning with dynamically weighted pseudolabeling provides a robust approach for prioritizing potential ADRs under severe data sparsity.
2 May 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 →International journal of pharmaceutics
From permeability to prediction: evolving strategies for evaluating oral drug absorption
Oral drug administration, preferred for patient compliance and convenience, encounters significant bioavailability challenges arising from complex physicochemical, physiological, and formulation-related barriers. Traditional models for evaluating absorption exhibit limitations in physiological relevance and predictive accuracy. This review critically examines evolving methodologies for oral drug absorption assessment, from classical permeability assays to emerging predictive platforms, emphasizing integration of mechanistic research with predictive modeling to advance intelligent platforms and clinical translation. While classical permeability models including parallel artificial membrane permeability assay (PAMPA), Caco-2, Ussing chambers, and intestinal perfusion remain indispensable for early screening, their limited physiological relevance and inability to elucidate mechanistic pathways constrain predictive accuracy. Mechanistic approaches such as lymphatic transport studies and advanced tracking techniques using fluorescence, Single Photon Emission Computed Tomography (SPECT), or Positron Emission Tomography (PET) imaging reveal critical biological pathways but suffer from scalability challenges. Emerging platforms provide transformative potential, as organoid-on-a-chip offer physiologically relevant and high-throughput alternatives, while integrated computational modeling enables multiscale prediction, from molecular interactions to systemic pharmacokinetics. Future oral absorption assessment requires converging in vitro biomimetics, computational modeling, and Artificial intelligence (AI) within unified platforms. Achieving this requires standardized in vitro models, robust data integration, and alignment with regulatory to accelerate translational drug development.
12 Apr 2026
Read appraisal →Analytica chimica acta
Point-of-need and portable sensing strategies for quality and safety assessment of Traditional Chinese Medicine products: A critical analytical review.
Sensor technology has emerged as a transformative tool for point-of-need and portable quality control and safety assessment of Traditional Chinese Medicine (TCM) products. Although separation and detection technologies have improved, there is still a strong need to critically evaluate the entire analytical workflow from sample preparation to data interpretation to meet regulatory and routine monitoring requirements. This review critically evaluates the complete analytical workflow, from sample pretreatment to intelligent data interpretation, necessary to bridge the gap between laboratory prototypes and real-world deployment. Additionally, it provides a systematic summary of the mechanisms and applications of common types of sensors, utilising nanomaterials and molecular recognition elements to overcome the specific challenge of complex TCM matrices. A particular focus is placed on the emerging roles of Artificial Intelligence (AI) and Machine Learning (ML) in resolving high-dimensional spectral data and enhancing pattern recognition accuracy. Furthermore, we critically analyze current bottlenecks, including matrix interference, the lack of standardized validation protocols, and the reliability of portable devices in uncontrolled environments. Future directions are highlighted towards the development of self-validating, cloud-integrated (IoT), and eco-friendly sensing systems, aiming to support the global standardization and safety assurance of TCM products.
9 Apr 2026
Read appraisal →JMIR mental health
It Is the Journey, Not the Destination: Moving From End Points to Trajectories When Assessing Chatbot Mental Health Safety
Large language models are rapidly becoming embedded in everyday life through artificial intelligence (AI) chatbots that people use for practical assistance and companionship, as well as for support with mental health and emotional well-being. Alongside clear benefits, clinicians and public reports increasingly describe a minority of users whose interactions seem to drift over days or weeks toward strongly questionable convictions, delusions, or suicidal crises. Importantly, clinically meaningful deterioration can occur even without overtly unsafe text outputs, via more insidious processes, such as compulsive use, sleep disruption, withdrawal from human contact, and progressive narrowing of attention around the chatbot relationship. In this Viewpoint, we argue that risk often arises not at a single tipping point but through trajectory effects that accumulate across extended dialogue and that prevailing safety evaluation approaches are misaligned with this reality because they primarily score risk at discrete conversational end points often reached through scripted dialogues lasting just a single turn or several turns. Mental health benchmarks and safety suites (including clinician-informed efforts) have advanced the field by testing refusal behavior, toxicity, and adversarial prompting. However, they often treat the last message as the unit of analysis and, therefore, miss when risk-relevant relational cues, signs of validation, contradiction handling, and shifts in certainty first emerge and how they compound. We propose that mental health safety assessment should shift from end points to trajectories by (1) treating the whole dialogue, not just the end result, as the focus of evaluation; (2) reporting turn-by-turn dynamics, such as delusion confirmation and harm enablement, and timing and persistence of safety interventions; and (3) calibrating short multiturn tests against longer, clinically realistic interaction sequences that can reveal context-length effects and drift. We further argue that transcript-only evaluation is insufficient in mental health contexts. Similar language can reflect very different internal states, and the relationship between expressed psychopathology and real-world harm is nonlinear. Therefore, safety research should incorporate proximal human outcomes following interactions (eg, shifts in certainty, openness to counterevidence, arousal, urge to continue, and subsequent sleep or behavior) and build a prospective clinical surveillance infrastructure that supports transcript donation with consent and linkage to health outcomes. Together, these steps would enable benchmarks that are clinically relevant and better aligned with the types of harms now being observed in real-world chatbot use.
7 Apr 2026
Read appraisal →