Browse Articles

Motivational regulations for exercise in Brazilian outpatients with bipolar disorder: An analysis based on the self-regulation theory

Despite the well-known physical activity (PA) benefits for physical and mental health, bipolar patients seem to be insufficiently active and often encounter challenges in participating in and adhering to programs involving PA. The present study aimed to assess the motivational regulation for exercise in bipolar disorder patients. This cross-sectional study utilizes objective (accelerometers) measures to assess PA. The Young Mania Rating Scale (YMRS) and the Hamilton Depression Rating Scale (HAM-D) were used to assess mania and depressive symptoms. The Behavioral Regulation in Exercise Questionnaire – 3 (BREQ-3) was used to assess the motivators for exercise. The sample was composed of 43 patients with bipolar disorder (81.5% female, mean age = 47 years; SD = 10.4). Sufficiently actives patients according to accelerometer moderate-vigorous physical activity (MVPA) were significantly more intrinsically regulated than those insufficiently actives (p = 0.002). The linear regression model showed that intrinsic regulation predicted 20% of the variability for accelerometer MVPA (p = 0.004). Notable contrasts were observed in integrated regulation. When asked whether they consider exercise as a part of their identity, 27.6% of the insufficiently active patients selected “Very true for me.” Meanwhile this view was endorsed by 57.1% of sufficiently active. Similarly, 50% of the active group reported that exercise is a fundamental part of who they are compared with 31% of the insufficiently active group. Our findings suggest that more autonomous forms of motivation, particularly integrated and intrinsic regulation, were associated with higher levels of PA among individuals with BD, whereas less active participants tended to present higher depressive symptomatology and lower motivational quality.

Interpretable classification of time series using Euler characteristic surfaces

Observed rapid adjustment of the atmospheric boundary layer to submesoscale sea surface temperature fronts

Current understanding of the role of ocean variability in air–sea exchange is constrained to large and mesoscale dynamics. Oceanic fronts and filaments with horizontal spatial scales of order 0.1 to 10 km—denoted submesoscale—are challenging to observe due to their fast-evolving flow and small spatiotemporal scales of variability. Observations investigating the air–sea fluxes at the submesoscale have shown substantial fluxes of heat, moisture, and momentum, affecting the structure of the overlying atmosphere. Here, modulations of the turbulent atmospheric boundary layer driven by ocean temperature anomalies are investigated using submesoscale-resolving ship and airborne measurements, providing in situ evidence of the atmospheric response to ocean submesoscale temperature variability. Observations suggest near-surface turbulent mixing driven by strong air–sea fluxes of heat and momentum, modifying the vertical structure of the planetary boundary layer. Linear regression coefficients between wind speed and sea surface temperature anomalies reveal a response similar in magnitude to that seen at larger scales, with an integrated change of 0.23 m s −1 °C −1 , but occurring over smaller length-scales, implying sharper gradients. Lagged correlations and scaling analysis imply a combined influence of horizontal advection and vertical turbulent mixing of momentum in the atmosphere, previously only described by numerical simulations. Observed cross-frontal wind divergences over the lower 200 m suggest coherent circulations with vertical velocities of order 1 cm s −1 . These observations confirm the rapid adjustment of the marine boundary layer to submesoscale ocean temperature variability and the importance of submesoscale-driven air–sea fluxes in changing the properties of the lower atmosphere, processes not resolved in most forecasting and prediction models.

Interlinked relationship between e-cigarette use and physical activity behaviour among Malaysian university students who use e-cigarettes: A cross-sectional study

Introduction E-cigarette (EC) use is increasing among young adults in Malaysia. However, evidence on how EC-related behaviours and perceived physical activity barriers are associated with physical activity participation remains limited. This study examined the associations between EC-related behaviours and perceived physical activity barriers with physical activity status among Malaysian university students who use EC. Methods A cross-sectional online survey was conducted between December 2023 and July 2024 across six Malaysian universities. Of 660 respondents, 564 met the age and International Physical Activity Questionnaire (IPAQ) criteria and were included in the analysis. Physical activity was classified as active and inactive. Exposures included EC use frequency, dual use, nicotine content, nicotine dependence, EC knowledge, motivations and perceptions for EC use, EC-related side effects, and perceived physical activity barrier items across personal, social and physical domains. Associations were analysed using chi-square tests and binary logistic regression, with odds ratios (ORs) and 95% confidence intervals. Results Overall, 64.9% were classified as active. At the crude level, EC use frequency, nicotine content, and dual-use status were associated with physical activity. However, these variables were not retained in the multivariable model due to multicollinearity. Nicotine dependence emerged as one of the strongest behavioural correlates of physical inactivity, with higher dependence associated with greater odds of physical inactivity. Perceived barriers to physical activity, particularly personal and social barriers, demonstrated the strongest and most consistent associations, with substantially higher odds of inactivity across barrier levels. EC knowledge and perception variables were not independently associated after adjustment, although item-level patterns were observed. Reported EC related symptoms were mainly gastrointestinal, respiratory, and neurological but were not analysed in relation to physical activity. Conclusion Physical activity among Malaysian university students who use EC is more consistently associated with nicotine dependence and perceived barriers, than with sociodemographic or knowledge variables. These findings suggest that interventions may benefit from addressing behavioural dependence and contextual constraints. Given the cross-sectional design, these results should be interpreted as hypothesis-generating. Implications Physical activity among university students who use EC appears more closely associated with nicotine dependence and perceived barriers than with knowledge or sociodemographic factors. While some EC use patterns showed crude associations, nicotine dependence emerged as one of the strongest behavioural correlates of physical inactivity. Lower perceived barriers, particularly personal and social, were consistently linked to higher activity participation. These findings suggest that intervention strategies may benefit from addressing behavioural dependence alongside reducing personal and social barriers to physical activity, particularly factors related to motivation, self-confidence, and competing demands. Integrating accessible, context-specific physical activity opportunities within EC cessation or harm-reduction programmes may enhance engagement and promote healthier lifestyle behaviours among young adults.

Cost-effectiveness of adding Helicobacter pylori screening to the national gastric cancer screening program in Korea

The ARHGAP32 isoform PX-RICS is specifically targeted to inhibitory synapses by binding to gephyrin

Precise regulation of excitatory–inhibitory balance is critical for neural circuit function, and its disruption underlies neurodevelopmental disorders such as autism spectrum disorder (ASD) and epilepsy. PX-RICS, a major ARHGAP32 splice variant enriched at inhibitory synapses, has been linked to cognitive dysfunctions; however, the molecular basis of its synaptic targeting and function remains unknown. Here, we identify gephyrin as the primary synaptic anchor for PX-RICS and determine the 2.2 Å crystal structure of their complex. Our structural analysis reveals that the N-terminal gephyrin-binding region (GBR) engages gephyrin E-domain through conserved hydrophobic interactions, explaining the isoform-specific targeting of PX-RICS (but not RICS) to inhibitory synapses. This binding interface overlaps with the neurotransmitter receptor binding site on gephyrin, suggesting a competitive yet dynamic interaction landscape among these inhibitory synaptic proteins. Arhgap32 ΔGBR mice exhibit key features of ARHGAP32 -related disorders, including impaired social novelty recognition and increased seizure susceptibility, indicating that gephyrin-mediated anchoring is critical for PX-RICS to function in inhibitory synapses.

Very short-term production prediction for photovoltaic plants using Temporal Convolutional Networks

Very short-term forecasting of solar photovoltaic energy production at national scale is challenging due to the high variability and spatial aggregation of generation across large territories. This paper evaluates Temporal Convolutional Networks (TCN) — a deep learning architecture based on causal and dilated convolutions — for nowcasting national-level solar production at one-hour and fifteen-minute horizons, using data from Spain sourced from the European Network of Transmission System Operators for Electricity. Multivariate models augmented with past weather observations (solar irradiance and sun height) are compared against linear regression baselines. Results demonstrate that the multivariate TCN substantially outperforms linear regression at the one-hour horizon, and achieves consistent improvement at the fifteen-minute horizon. The relative contribution of architecture and weather features is resolution-dependent: at hourly granularity, the TCN architecture itself provides the dominant gain, while at the fifteen-minute scale the inclusion of weather covariates becomes the primary driver of accuracy, reflecting the greater atmospheric variability at finer temporal scales. A key finding is that past-only weather inputs are sufficient for accurate nowcasting, eliminating the need for future meteorological forecasts as model inputs. The results support the practical applicability of TCN-based models for national-level solar energy integration and provide a data-driven feature-selection criterion for similar renewable energy forecasting tasks.

A hybrid experimental and machine learning framework for designing and predicting compressive strength of ultra-high-performance concrete

Abstract Ultra-high-performance concrete (UHPC) offers exceptional mechanical and durability properties but often relies on quartz powder, raising sustainability and occupational health concerns. This study introduces an integrated experimental-computational framework for predicting the compressive strength of UHPC and developing quartz-free mixtures. Experimentally, the effects of mixing sequence, sand characteristics, superplasticizer chemistry, and curing regime were investigated, leading to a quartz-free UHPC achieving 136 MPa at 28 days under heat-curing. A dataset of 550 UHPC compressive strength records was compiled, incorporating quantitative mix proportions and categorical variables (cement type, superplasticizer base, fiber type, and specimen geometry). Sixty-three machine learning models from tree-based, boosting, and support vector machine families were optimized using seven meta-heuristic algorithms. The Particle Swarm Optimization-tuned XGBoost model achieved the highest prediction accuracy (R 2  = 0.897, RMSE = 7.63 MPa), followed by the Differential Evolution-optimized Random Forest (R 2  = 0.867, RMSE = 8.70 MPa). SHapley Additive exPlanations (SHAP) analysis identified curing age as the most influential predictor after optimization. The proposed framework enables accurate and interpretable UHPC strength prediction and supports the design of safer and more sustainable quartz-free UHPC with reduced experimental effort.

H ₂ S-mediated protein persulfidation regulates redox metabolic flux underlying salt-stress resilience in rice

Hydrogen sulfide (H 2 S) functions as a gaseous signaling molecule in plant stress responses through the persulfidation of protein cysteine (Cys) residues. A comprehensive, Cys site-specific map of the plant persulfidome has been lacking, despite its importance for achieving a systems-level understanding of the biological roles of Cys persulfidation. Using a state-of-the-art N -ethylmaleimide-biotin-based proteomics strategy, we generate a dynamic map of 1,691 persulfidated Cys sites in the rice ( Oryza sativa ) leaf proteome. Our results reveal a global dynamic changes in protein persulfidation during prolonged salt stress, with notable impacts on proteins involved in metal-dependent catalysis, redox metabolism, and the pentose phosphate pathway (PPP). Based on these patterns, we investigated the functional relevance of persulfidation within the nonoxidative PPP. H 2 S-mediated persulfidation decreased the activity of the representative nonoxidative PPP enzyme ribose-5-phosphate isomerase, leading to increased NADPH production and subsequent activation of NADPH-dependent redox enzymes, including monodehydroascorbate reductase (MDHAR) isoforms of the ascorbate-glutathione (AsA-GSH) cycle. Persulfidation protected MDHAR3/5 from oxidative inhibition and degradation, thereby sustaining AsA-GSH cycle capacity and supporting reactive oxygen species scavenging. This site-specific persulfidome provides a valuable resource for exploring plant redox regulation, and our functional analyses offer mechanistic insight into how H 2 S-dependent protein persulfidation modulates redox metabolic fluxes to bolster NADPH availability and maintain redox homeostasis during salt-stress adaptation.

Bisphenol A exposure in myasthenia gravis: Potential targets and mechanisms revealed by network toxicology and molecular dynamics

Background Myasthenia gravis (MG) is a B-cell-mediated autoimmune disease characterized by impaired neuromuscular transmission. Although genetic predisposition and thymic abnormalities are well recognized, they cannot fully explain the increasing incidence and regional heterogeneity of MG, highlighting the potential contribution of environmental factors. Bisphenol A (BPA), a ubiquitous endocrine-disrupting chemical, exhibits estrogenic activity, immunomodulatory effects, and mitochondrial toxicity, and has been implicated in multiple autoimmune disorders. However, the potential role of BPA in MG pathogenesis remains largely unexplored. Methods BPA-related targets and MG-associated genes were collected from public databases, and overlapping targets were identified. Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses were performed to explore the biological functions and pathways of the overlapping targets. Candidate targets were screened using four machine-learning algorithms, including least absolute shrinkage and selection operator (LASSO) regression, support vector machine-recursive feature elimination (SVM-RFE), random forest (RF), and extreme gradient boosting (XGBoost). Differential expression and diagnostic performance were validated using the Gene Expression Omnibus (GEO) dataset GSE85452. Immune infiltration was assessed using CIBERSORT. Molecular docking and molecular dynamics (MD) simulations were conducted to evaluate the potential binding stability and interaction modes between BPA and key target proteins. In addition, C2C12 myoblasts were treated with different concentrations of BPA for 24 h, and cell viability was assessed using the Cell Counting Kit-8 (CCK-8) assay. Based on the cell viability results, 50 μM BPA was selected for quantitative reverse transcription polymerase chain reaction (qRT-PCR) and Western blot analyses of key genes. Results A total of 225 overlapping targets were identified between BPA exposure-related targets and MG-associated genes. Enrichment analyses showed that these targets were mainly associated with ion transport, membrane potential regulation, muscle system processes, immune signaling, apoptosis, endocrine resistance, and AGE–RAGE signaling pathways.Four machine-learning algorithms identified cholinergic receptor nicotinic beta 1 subunit (CHRNB1), KRAS proto-oncogene, GTPase (KRAS), phosphomannomutase 2 (PMM2), and toll-like receptor 4 (TLR4) as candidate targets. Validation in the GSE85452 dataset showed that CHRNB1, PMM2, and TLR4 were significantly upregulated in MG samples compared with control samples, whereas KRAS showed no significant difference. receiver operating characteristic (ROC) analysis further demonstrated that CHRNB1, PMM2, and TLR4 had good diagnostic performance, with area under the ROC curve (AUC) values of 0.827, 0.856, and 0.821,respectively. Immune infiltration analysis revealed altered immune cell infiltration patterns and associations between key targets and specific immune cell subsets. Molecular docking predicted favorable binding of BPA to CHRNB1, PMM2, and TLR4, with binding energies of −7.4, −5.7, and −5.8 kcal/mol, respectively. MD simulations further supported the potential stability of these BPA-target complexes. In vitro experiments showed that BPA reduced C2C12 cell viability in a concentration-dependent manner. qRT-PCR validation showed that treatment with 50 μM BPA significantly upregulated Chrnb1 expression, while downregulating Pmm2 and Tlr4 expression.Western blot analysis further confirmed that BPA exposure significantly decreased the protein expression of TLR4 and PMM2, while increasing that of CHRNB1 in C2C12 cells. Conclusions This study provides integrated computational and experimental evidence that BPA exposure may be associated with MG-related molecular alterations. BPA may affect MG-related biological processes through the regulation of ion transport, neuromuscular signaling, immune activation, and glycosylation-related metabolism. CHRNB1, PMM2, and TLR4 may serve as potential molecular links between BPA exposure and MG-related pathological processes.

Tunable broadband multi-functional perovskite graphene-based polarization convertor at THz band for 6G applications

Children’s books about scientists convey demotivating messages

Biographies of scientists are widely used to spark children’s interest and broaden participation in science. In telling the stories of successful scientists, these books convey implicit “recipes for success”—yet no study has systematically examined whether these recipes align with what motivation research recommends. Here, we provide the first large-scale analysis of motivational messages in children’s science biographies (422 best-selling books, 1,355 protagonists). Four main findings emerged. First, scientific ability was portrayed equally often as fixed and as malleable, a pattern unlikely to support motivation. Second, interest in science was overwhelmingly portrayed as fixed: Most scientists were said to be captivated by science from a young age. Third, scientists in physics, engineering, and computer science were portrayed as more reliant on fixed ability than scientists in other fields. Fourth, biographies of female scientists disproportionately emphasized effort even after adjusting for the obstacles they faced, consistent with stereotypes attributing women’s success to hard work over talent. These findings suggest that well-intentioned efforts to diversify science through role model exposure in children’s science biographies may fall short.

Prognostic factors in first-line atezolizumab-bevacizumab treatment of intermediate or advanced hepatocellular carcinoma

Purpose To analyze the prognostic factors influencing outcomes in hepatocellular carcinoma (HCC) patients initially treated with atezolizumab-bevacizumab (Ate-Bev). Methods We conducted a retrospective study on 47 treatment-naïve HCC patients (40 men; median age, 56 years) who were treated with Ate-Bev between September 2020 and July 2022 and underwent at least one response assessment based on the Response Evaluation Criteria in Solid Tumors 1.1. We evaluated the clinical variables and imaging features of pre-treatment magnetic resonance imaging (MRI). Using multivariable Cox hazard regression analysis, we analyzed prognostic factors for progression-free survival (PFS). Results During the follow-up period (range, 1.5–24.9 months), 33 patients (70.2%) experienced tumor progression. The median PFS was 4.3 months (95% confidence interval [CI], 2.8–8.5 months). Nine patients received concomitant radiotherapy. Multivariable analysis indicated that younger age (hazard ratio [HR], 0.95; 95% CI, 0.92–0.99; P  = 0.013), absence of concomitant radiation therapy (HR, 0.22; 95% CI, 0.07–0.69; P  = 0.009), tumor extent ≥10 cm (HR, 2.54; 95% CI, 1.17–5.55; P  = 0.019), and peritumoral arterial-phase hyperenhancement on the MRI (HR, 2.15; 95% CI, 1.00–4.63; P  = 0.049) were factors associated with poor PFS. Conclusions Age, concomitant radiation therapy, tumor extent, and peritumoral arterial-phase hyperenhancement on MRI were associated with PFS in patients with HCC who were initially treated with Ate-Bev.

Multiscale acoustic temporal niche partitioning of pineland birds across a species richness gradient

Abstract Growing evidence suggests birds reduce acoustic competition by minimizing temporal overlap with neighboring species sharing similar vocal characteristics. However, few studies have tested this across ecological contexts at multiple temporal scales. We examined temporal vocalization patterns of five common pine woodland bird species across a species richness gradient in Florida, USA. We predicted that Carolina Wren ( Throthorus ludovicianus) and White-eyed Vireo ( Vireo griseus ) would exhibit temporal avoidance with the acoustically similar, highly abundant and vocal Northern Cardinal ( Cardinalis cardinalis) at fine (1-minute) and coarse (3-hour dawn chorus) temporal scales, with stronger avoidance in species-rich communities. Conversely, we predicted acoustically dissimilar species [Mourning Dove ( Zenaida macroura ) and Pine Warbler ( Setophaga pinus )] would show random temporal associations with Northern Cardinals, unaffected by richness or scale. Consistent with predictions, acoustically similar species exhibited significant temporal avoidance of Cardinals at both scales, while dissimilar species showed no consistent patterning. Contrary to expectations, species richness did not strongly influence temporal avoidance, though fine-scale temporal partitioning trended toward increasing at lower richness. Taken together, these results demonstrate that temporal niche partitioning among acoustically similar bird species operates consistently across multiple temporal scales and is structured more by acoustic similarity and asymmetric species interactions than by species richness.

ZNF263–NuRD-mediated repression of STAT1 curtails MHC-I antigen presentation and IFN-γ efficacy in prostate cancer

Loss of major histocompatibility complex (MHC)-I is a hallmark of prostate cancer (PCa) immune evasion and immunotherapy failure. Here, we identify ZNF263 as a transcriptional repressor that silences MHC-I by recruiting nucleosome-remodeling and deacetylase (NuRD) to the STAT1 promoter, reducing STAT1 and MHC-I expression. Hypoxia enhances this repression through two ZNF263 modifications: phosphorylation-driven phase separation that strengthens NuRD interaction and O-GlcNAcylation at S662 that aids STAT1 promoter binding. O-GlcNAcylation also promotes interaction with protein kinase, DNA‑activated catalytic subunit (PRKDC), amplifying phosphorylation. Interferon‑gamma (IFN‑γ)‑induced MHC-I induction is augmented upon ZNF263 loss. In silico docking identified Viroptic as a Krüppel‑associated box (KRAB) pocket binder disrupting ZNF263–NuRD, derepressing STAT1, and potentiating IFN-γ antitumor immunity in vivo. High ZNF263 correlates with low MHC-I, scarce CD8+ T cells, and poor survival, providing rationale for targeting ZNF263 in PCa immunotherapy.

New insights into the Trans-Saharan gold trade in the Islamic Middle Ages revealed by multi-isotopic (Pb–Fe–Cu) and elemental characterization of Fatimid gold coins

The use of West African gold in medieval times is thought to have enabled the first globalized trade in history: the Trans-Saharan Trade, through the minting of Islamic dinars, which were considered as the equivalent of today’s dollar or euro currencies. Yet, the precise origin of this gold remains debated, as archaeological evidence for such trade is scarce. While textual sources are numerous, they provide only indirect accounts, and the large corpus of surviving dinars has so far yielded limited insight as elemental analyses are not discriminating enough to establish geological provenance. To address this, we combined multi-elemental analysis by LA-ICP-MS with multi-isotope (Pb, Cu, and Fe) data measured by MC-ICP-MS after wet purification chemistry on 11 dinars. This revealed two distinct metal stocks, differentiated by PGE concentrations and Cu isotopes. The Cu and Fe isotope signatures are consistent with supergene and oxidized ores, typical of West African gold mineralization. In contrast, the lead isotopes and calculated model ages of the gold coins do not match the signatures expected for West African mineralization, suggesting that a different, exogenous source is involved in the gold chaîne opératoire . These new data then raise two key historical questions: the location of gold processing, whether it occurred North or South of the Sahara, and the hypothetical reuse of older metal stocks.

Placental lipid handling, growth and inflammatory pathways are modified by a maternal Mediterranean diet

Abstract The Mediterranean diet is associated with reduced cardiometabolic risk, yet its physiological effects during pregnancy and its impact on placental metabolism remain incompletely understood. This study aimed to determine whether maternal adherence to a Mediterranean diet during pregnancy influences placental lipid metabolism and signalling pathways involved in nutrient handling, tissue remodelling, and inflammation, and to assess their relationship with pregnancy outcomes. Placental samples and clinical outcome data were analysed from pregnant women participating in an unblinded randomized clinical trial of a Mediterranean diet intervention. Placental fatty acid composition was quantified, and the expression of genes and signalling pathways involved in lipid metabolism, nutrient transport, inflammation, and tissue remodelling was evaluated. Maternal adherence to a Mediterranean diet during pregnancy was associated with significant alterations in placental fatty acid composition, including reduced C18:0 and C24:0 and increased C18:1n9c, C20:3n6, and C22:0, with lower total saturated fatty acids and higher monounsaturated fatty acids. Placental expression of lipid metabolism regulators ALOX15 and PPARγ was reduced, alongside downregulation of AKT and p38 MAPK signalling pathways. Placentas from mothers adhering to the Mediterranean diet also showed lower expression of amino acid and glucose transporters SLC3A2 and SLC2A1 , as well as altered inflammatory and extracellular matrix remodelling markers, including decreased SOCS3 and GHR and increased PAI1 and MMP3. Maternal adherence to a Mediterranean diet during pregnancy modifies placental fatty acid composition and regulates pathways involved in lipid handling, nutrient transport, inflammation, and tissue remodelling, providing insight into mechanisms linking maternal diet with placental metabolic function.

A minimal wake–vortex model explains formation flight of flapping birds

Collective patterns of motion emerge across biological taxa: insects swarm, fish school, and birds flock. In particular, many large migratory bird species form distinctly ordered V-shaped formations, which experiments and direct numerical simulations have demonstrated provide substantial energetic benefits during long-distance flight. However, the precise aerodynamic and morphological mechanisms which underlie these benefits remain unclear. In this work, we develop a reduced-order model of the wake–vortex interactions between two flapping birds flying in tandem. The model retains essential unsteady flapping dynamics while remaining computationally tractable. By optimizing over a six-dimensional state space, which comprises the follower’s three-dimensional relative position as well as three independent flapping parameters, we identify the energetically optimal leader–follower configuration of northern bald ibises ( Geronticus eremita ). The predicted optimum agrees quantitatively with live-bird measurements. Because of its simplicity, the model allows for direct interrogation of the physical mechanisms responsible for this optimum. In particular, it isolates precisely how the follower’s wing kinematics interact with the leader’s wake to enhance aerodynamic efficiency. The model predicts an 11% reduction in total mechanical power for a follower in formation flight—consistent with experimental estimates—and shows that this saving arises from reductions in both induced and profile power, dominated by decreased profile power enabled primarily through reduced flapping amplitude and, secondarily, reduced upstroke flexion. These results provide a mechanistic explanation for the structure of V-formations and offer insight into the aerodynamic principles governing collective flight.

Early-warning signals for infectious diseases with a social-media compartment

Early-warning signals (EWSs) are crucial tools for anticipating disease emergence and guiding public-health responses, but uncertainties in transmission and incomplete data can limit their reliability. Additionally, the performance of EWSs is rarely evaluated when disease emergence is delayed, and their use in the context of interactions between disease transmission and communication through social-media platforms has largely not been considered. We evaluate the relative performance of EWSs in predicting disease emergence under varying noise conditions and explore the use of EWSs with social-media dynamics to predict disease emergence. We develop a mechanistic model coupling infectious disease and social-media dynamics, introduce stochasticity and generate simulated time series. We detect changepoints, quantify delays relative to the bifurcation point and assess the performance of EWSs across different segments of the time series. The “reporting” infected compartment proves most reliable, and variance outperforms autocorrelation in high-noise scenarios. However, in the social-media compartment, variance and autocorrelation have weak predictive power. Our work provides a framework to advance understanding of how EWSs can be applied to forecasting disease emergence, contributing to improved disease preparedness and response.

Domain-specific versus general large language models: a review and empirical benchmark in real medical texts

Abstract The rapid adoption of Large Language Models (LLMs) has led to their widespread use as a general-purpose, fast-deployment solution across a broad range of tasks. In healthcare settings, this trend has accelerated the replacement of task-specific models with large generative architectures, often without an assessment of their efficiency or suitability for structured information extraction. While transformer-based encoder models have long been established as reliable solutions for clinical Named Entity Recognition (NER) and Information Extraction (IE), the increasing reliance on LLMs raises concerns regarding computational cost, scalability, energy consumption, and practical deployment in resource- and privacy-constrained environments. This study aimed to critically evaluate the assumption that LLMs constitute a universally superior solution for clinical NER and IE, by systematically comparing encoder-based Language Models (LMs), and LLMs in terms of extraction performance, efficiency, deployment feasibility, and environmental impact. Following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA 2020) guidelines, a scoping literature review was conducted using Scopus and Web of Science, covering peer-reviewed studies published between 2022 and 2025. In addition, a proof-of-concept case study was performed on European Portuguese clinical notes, comparing a domain-adapted encoder model (MediAlbertina-1.5B) with quantized Llama-based LLMs executed locally under identical hardware conditions. Models were evaluated using micro-averaged precision, recall, and F1-score, together with runtime, estimated energy consumption, and CO₂ emissions. The literature review showed that LLMs and transformer-based architectures were the most frequently used even thought encoder-based and domain-adapted models often perform strongly in clinical NER, particularly when recall, structured output, and computational efficiency are important. The empirical case study showed a similar pattern: MediAlbertina-1.5B achieved the highest performance (micro-F1 = 0.430), whereas all evaluated Llama variants obtained markedly lower F1-scores (≤ 0.123) due to systematic under-extraction. Despite their lower performance, the Llama models incurred 23–95 × longer runtimes and one to two orders of magnitude higher estimated energy consumption and CO₂ emissions under the tested local CPU-only setup. In this proof-of-concept benchmark on 20 European Portuguese clinical reports, the domain-adapted MediAlbertina token-classification model achieved higher recall and micro-F1 than the evaluated locally executed, quantized Llama models. The Llama models were also substantially slower and less sustainable, with higher estimated energy use and CO₂ emissions under the tested hardware and prompting configuration. These findings suggest that domain-adapted encoder models may be preferable for similar structured clinical NER settings, although larger and externally validated benchmarks are needed before generalizing to other languages, clinical corpora, model families, or deployment environments.