Most diseases result from complex molecular interactions of genes and proteins. Various network-based methods characterize these mechanisms by expanding seed genes into disease modules. Their underlying algorithmic strategies differ, making it difficult to determine which of the created modules are most useful or biologically plausible. To address this challenge, we developed an all-in-one pipeline that handles installation, input preparation, execution, and systematic evaluation of six widely used module detection tools, considering module topology, functional coherence, robustness, and the capacity to recover seeds. To showcase the value of our pipeline and provide guidance to potential users, we conducted a comprehensive evaluation across 50 different disease-network combinations, revealing substantial variability among the derived disease modules, driven by both network and algorithm choices. We show that methods are robust to minor perturbations but struggle to recover omitted seeds. None consistently outperforms all others, underscoring the need for careful method selection. Our work enables the systematic comparison of disease module discovery approaches and promotes reproducible network medicine research. Integrated into the nf-core project, it is intended as an extendable, long-term resource for tracking progress in the field. The pipeline is implemented in Nextflow. Code and documentation are available through GitHub (https://github.com/nf-core/diseasemodulediscovery) and the nf-core website (https://nf-co.re/diseasemodulediscovery). Code and data used for demonstrating the pipeline are available through GitHub (https://github.com/REPO4EU/modulediscovery_demonstration).
Increasing health concerns and awareness among consumers have driven a significant shift toward healthier food choices, fuelling demand for functional foods that are minimally processed and rich in bioactive compounds. Exploiting microbial co-cultivation that can trigger quorum sensing shows tremendous potential for synthesising valuable functional food compounds. This study explored the influence of co-cultivation between two lactic acid bacteria (LAB), Lactococcus lactis (LL) and Lacticaseibacillus rhamnosus (LR) with various fermentation parameters (temperature, pH, agitation, carbon, and nitrogen sources) to determine its potential for enhancing the production of riboflavin and nisin. The highest riboflavin production (40.6 ng/mL) was attained in co-culture at 37°C with 100 rpm agitation, while the highest nisin activities for co-cultivation against Listeria monocytogenes (11.5 mm) and Bacillus subtilis (8.8 mm) were at 200 and 100 rpm, respectively. This study provides a fundmental exploration of the interspecies dynamics between L. lactis and L. rhamnosus through one-factor-at-a-time (OFAT) approach in shake-flask fermentations. These findings serve as a preliminary proof-of-concept for co-cultivation efficiency that could improve targeted compound production.
The Lorentz factor is a fundamental correction factor in quantitative analysis of X-ray diffraction experiments, enabling measured integrated peak intensities to be related to calculated structure factors. In this review, the physical origin of the Lorentz factor, its derivation and practical implementations are presented based on a unified approach, treating the Lorentz factor as the Jacobian relating the experimental measurement coordinates to the reciprocal-space volume element. This approach clarifies the trigonometric and wavelength-dependent contributions to the Lorentz factor and explains the origin of the commonly used `angular-velocity' formulation for rotational measurements. Equations for the Lorentz factor are derived systematically for a broad range of modern X-ray diffraction geometries, including single-crystal and powder diffraction, grazing-incidence diffraction methods, as well as small-angle X-ray scattering. Hereby it is demonstrated that the Lorentz factor sensitively depends not only on the experimental geometry, but also on the type of sample under investigation. In addition, we discuss how the Lorentz correction may be avoided by direct reciprocal-space integration, and the practical limitations of such approaches are pointed out. By providing a unified and comprehensive overview of the Lorentz factor, this review aims to support reliable and consistent intensity evaluation across various modern X-ray diffraction methods.
Sequential multiple assignment randomized trials (SMARTs) mimic the actual treatment processes experienced by physicians and patients in clinical settings and inform the comparative effectiveness of dynamic treatment regimes. In such trials, patients undergo multiple stages of treatment, and the treatment assignment is adjusted over time based on individual patient characteristics, such as disease status and treatment history. In this work, we develop and evaluate statistically valid global interim monitoring (IM) approaches to allow for early termination of SMARTs for efficacy targeting survival outcomes. We propose a weighted log-rank Chi-square statistic to account for overlapping treatment paths and quantify how the log-rank statistics at two different analysis points are correlated. Efficacy boundaries at multiple interim analyses can then be established using the Pocock, O'Brien Fleming, and Lan-DeMets boundaries. We conduct extensive simulations to comparatively evaluate the operating characteristics (type I error and power) of our IM procedure based on the proposed statistic and an existing alternative statistic. The methods are demonstrated via an analysis of a neuroblastoma dataset.
Diagnostic Reference Levels (DRLs) are a fundamental instrument for optimizing patient radiation protection in medical imaging. However, establishing and sustaining a nationally representative DRL system in a geographically vast and resource-diverse country presents significant methodological, regulatory, and operational challenges. Ensuring nationwide dose reporting, harmonizing heterogeneous examination protocols, and translating benchmark values into sustained clinical optimization remain complex tasks, particularly in large archipelagic settings. This study presents Indonesia's experience in establishing, implementing, and periodically reviewing its national DRLs across all major diagnostic modalities through a regulator-led national dose optimization framework. The Indonesian DRLs was established through a nationwide dose survey using the National Patient Dose Data Information System, enabling cumulative nationwide dose data collection from more than 1000 licensed healthcare facilities between 2015 and 2025. Dose data were obtained using direct and indirect dosimetry methods, and Indonesian DRL (IDRL) values were derived as the 75th percentile of facility median doses following regulator-led multi-stakeholder validation. Indonesian DRLs have been established for general radiography, computed tomography (CT), interventional fluoroscopy, mammography, dental radiography, and diagnostic nuclear medicine. The IDRL is subject to periodic review at least every five years, or earlier in response to significant technological advancements or changes in clinical practice. The 2025 review showed reduced DRL values in radiography and CT, indicating improved optimization practices, whereas increases in interventional fluoroscopy reflected growing procedural complexity and the need for complexity-adjusted approaches. Key challenges include variable facility participation, heterogeneous examination protocols, gaps in dosimetry expertise, and a shortage of medical physicists. Indonesia's experiences highlight that sustainable DRL implementation requires robust data systems, stakeholder engagement, regulatory enforcement, and capacity building. This experience offers valuable insights for other large and diverse countries seeking to establish effective national DRL frameworks.
While ferromagnetism is well understood in periodic and amorphous materials, its critical behavior in quasiperiodic systems has remained elusive. Although ferromagnetic icosahedral quasicrystals (iQCs) have recently been discovered, all known examples have been limited to rapidly quenched, metastable states, precluding annealing, structural refinement, and quantitative studies of magnetic criticality. Here we demonstrate, for the first time, the realization of bulk, annealable ferromagnetic iQCs in compositionally tuned Au-Cu-Al-In-R (R = Gd, Tb, Dy) alloys. Guided by chemical design in multicomponent alloy space, single-phase iQCs are obtained by conventional arc melting and subsequent controlled annealing, resulting in sharp quasiperiodic diffraction and exceptional thermal robustness. Strikingly, while all compounds exhibit clear ferromagnetic order, their magnetic critical behavior differs systematically depending on the rare-earth element. Tb- and Dy-based iQCs display critical exponents close to mean-field values, whereas the Gd-based system exhibits a significantly enhanced critical exponent δ, deviating from mean-field behavior and from both anisotropic counterparts and previously studied approximant crystals. This difference is attributed to stronger spin fluctuations in the isotropic Gd system, which effectively shorten the interaction range, while anisotropy in Tb/Dy suppresses fluctuations and leads to mean-field-like behavior. These results establish that magnetic criticality in quasicrystals is not uniquely determined by quasiperiodic order alone but is governed by the interplay between quasiperiodicity and local spin symmetry. More broadly, they demonstrate that quasiperiodic solids provide a tunable materials platform for accessing nonmean-field critical behavior beyond the constraints of periodic systems.
In recent years, many countries have experienced increased outbreaks of vector-borne diseases, such as West Nile virus (WNV). WNV is primarily transmitted to humans by the bite of infected Culex mosquitoes and has an enzoonotic transmission cycle involving birds. In Algeria, serological evidence of WNV in humans dates back to the 1970s, but circulation patterns are poorly characterized. To address this, we conduct a cross-sectional seroprevalence survey of WNV in humans and fit serocatalytic models to the data. We also analyze data on severe WNV cases. In this study, we report a seroprevalence of 16.7%. We identify age, region, and residence type as risk factors for infection. We also show that patterns of spread in two provinces in the south are consistent with low level endemic risk over decades, while those from three provinces in the north are all consistent with single outbreaks that occurred within the last 10-15 years. An outbreak may have occurred in Oran in 2010 and in Jijel and Tizi Ouzou near the time of data collection in 2017 and 2018, respectively. Investigation of potential drivers of viral establishment, along with continued surveillance of WNV cases, especially in the northern provinces, is warranted.
Epigenetic clocks based on DNA methylation (DNAm) accurately predict age, but their biological underpinnings remain unclear. One primary mechanism by which DNAm might influence gene regulation is by modulating transcription factor binding activity. This study investigates the regulatory potential of predictive CpGs in established epigenetic clocks. Our analysis reveals that generally most CpGs used by epigenetic clocks do not overlap known transcription factor binding sites (TFBS), indicating that clock accuracy is not primarily driven by changes in TF binding dynamics. However, analysis of CpGs within TFBSs identifies key transcription factors potentially involved in aging, including ZBED1, NFE2, and CEBPB, which are enriched for age-associated CpGs, while RELA, IKZF1, and STAT3 significantly protected against methylation changes. Leveraging TFBS-associated and age-correlated CpGs, combined with noise-stabilizing feature engineering steps, we developed an alternative TFMethyl Clock model that provides competitive predictions of chronological age. Age-predictive CpGs selected by our model enrich for target genes involved in interleukin-1β production and fatty-acid metabolism, while being enriched at TFBSs of NR2C2. Furthermore, approximately three-fourths of these target genes exhibit significant age-related changes, suggesting deeper insights into possible methylation-driven aging processes. Our findings demonstrate that incorporating regulatory information into epigenetic clocks may provide mechanistic insights into the aging process while improving the interpretability and predictive power.
Organellar genomes are both a resource for reconstructing organismal phylogenies and interesting subjects for evolutionary studies. Herein, we focused on the organellar genomes of eustigmatophytes (eustigs), a class of the algal phylum Ochrophyta with a growing biotechnological potential, and massively expanded the existing limited sample by 51 new organellar genomes. Analyses of this large dataset provided a robustly resolved eustig phylogeny and important insights into the evolution of unique features of eustig organellar genomes. Eustig plastomes are rather stable in terms of the gene content, with only minor differences stemming from differential gene loss and rare lineage-specific gain. In contrast, eustig mitogenomes contain a very stable core of conserved genes accompanied by a broadly varying shell comprising "accessory" genes. Notably, the new data illuminated the origin of two mitochondrial genes previously deemed eustig-specific, namely orfX and orfY that were found to have evolved, respectively, by rps4 duplication and extreme divergence of an rps1 ortholog. Most interestingly, we identified five previously unrecognized orthogroups of mysterious mitochondrial orfs that are patchily distributed across eustigs yet likely evolved in the ancestor of this class. These orfs have no discernible homologs outside eustigmatophytes but are predicted to encode multipass membrane proteins with a soluble C-terminal domain. Our results also revise some of the previous conclusions regarding the mitochondrial translation in eustigs and suggest the recruitment of a group of unusual tRNAs for a translation-independent function in the genus Vischeria. Our study thus provides a glimpse into a "dark matter" of mitochondrial biology in eustigmatophytes.
This systematic review and meta-analysis was registered with PROSPERO (Registration No. CRD420251130849) and conducted in accordance with PRISMA 2020 guidelines. A comprehensive literature search of PubMed, Web of Science, and SciencDirect (2000-2025) identified English-language studies reporting the diagnostic performance of HPV DNA, HPV E6/E7 mRNA, and p16INK4a for cervical cancer. Studies providing sufficient data to calculate sensitivity and specificity were included. Pooled estimates were generated using a random-effects model,heterogeneity was assessed using the I2 statistic, and methodological quality was evaluated with the QUADAS-C tool. Overall, 77 studies involving 17,558 women were included. HPV DNA demonstrated a pooled sensitivity of 0.783(95% CI: 0.611-0.956) and specificity of 0.731 (95% CI: 0.584-0.877), with considerable heterogeneity (I2 = 88-96%). HPV E6/E7 mRNA showed the highest diagnostic accuracy, with a pooled sensitivity of 0.950 (95% CI: 0.930-0.970) and specificity of 0.972 (95% CI: 0.959-0.986). Although substantial heterogeneity was observed (I2 = 90%, p < 0.01), the biomarker demonstrated consistently superior diagnostic performance across studies. p16INK4a yielded a pooled sensitivity of 0.874 (95% CI: 0.758-0.989) and a specificity of 0.756 (95% CI: 0.56-0.950), with high heterogeneity (I2 = 92-96%), indicating marked inter-study variability. Among the evaluated biomarkers, HPVE6/E7 mRNA exhibited the highest diagnostic accuracy and sensitivity for cervical cancer detection in both Indian and U.S. populations. Nevertheless, despite its promising performance, larger, well-designed multicenter studies with standardized methodologies and external validation are needed before these biomarkers can be reliably implemented in routine clinical practice.
Endometriosis is a chronic, hormone-dependent condition affecting an estimated 190 million women worldwide. Our understanding of hormonal contributions to endometriosis pathophysiology is incomplete, hindering the identification of diagnostic biomarkers and novel therapeutic targets. Although the role of estrogens is well established, research on androgens in endometriosis is limited and the contribution of adrenal-derived 11-oxygenated androgens remains largely unknown. We performed steroid androgen profiling to measure androgen concentrations in serum from healthy controls and women with laparoscopically confirmed endometriosis. We found that women with endometriosis had a distinct hormone signature characterized by systemic differences in adrenal androgen concentrations and 11-ketotestosterone excess.Using metabolomic data, we generated statistical models that showed robust discrimination between healthy controls and women with endometriosis (AUC = 0.99; positive predictive power = 96.84%, negative predictive power = 92.86%) consistent with an endometriosis-specific signature. Data were partitioned into train and validation groups to assess diagnostic potential and a refined model identified >95% of endometriosis patients in a blinded sample set. Collectively, these data reframe endometriosis as an androgen-dependent disorder and highlight 11-oxygenated androgens as potential diagnostic biomarkers and future therapeutic targets.
The current research focuses on optimizing extraction parameters for isolating Ursolic acid from Plectranthus amboinicus (Indian balm), Rosmarinus officinalis (Rosemary leaves), and Ocimum basilicum (Basil leaves), using both maceration, a conventional technique as well as ultrasound-assisted extraction. A one-factor-at-a-time approach was employed to optimize extraction parameters. Among the three plant sources analyzed, rosemary exhibited the highest crude yield and ursolic acid content, and HPLC analysis confirmed the presence of ursolic acid with 26.956 mg/L in rosemary, 16.374 mg/L in basil, and 0.359 mg/L in Indian balm. Thus, rosemary was selected for further modeling and optimization to provide a basis for future scale-up studies, and mathematical models were applied to describe the effects of individual parameters, all showing good agreement of R2 > 0.96. RSM based optimization was further applied, which showed that ultrasonic power of 157.5 W at 45 °C for 40 min was the optimum condition with a maximum yield of 76.48%, with the model showing an R2 of 0.9990 and statistical significance (p < 0.0001). The optimized UAE process showed remarkable extraction efficiency compared to conventional maceration. Thus, this study, combining experimental optimization of extraction parameters, mathematical modeling, and RSM for efficient extraction of ursolic acid, is an integrated approach.
First passage phenomena arise across physics, biology, and finance when stochastic processes first reach a threshold, triggering downstream events. Examples include the irreversible exit from a domain, a biochemical reaction, and a financial selloff. While typical formulations involve diffusive motion, many stochastic processes are better described as velocity jump processes, characterized by persistent motion interrupted by stochastic velocity changes. Despite their ubiquity, first passage properties of velocity jump processes remain underdeveloped in higher dimensions, especially under directional bias. We present a general framework to estimate the mean first passage time (MFPT) and higher moments of the survival probability for fixed-speed velocity jump processes where possible reorientations range from strong alignment to full angular anisotropy. For low Knudsen numbers, when the mean free path is small compared to the distance to the target, we derive a universal form for the MFPT in which two bias functions encode broad classes of angular distributions, including von Mises-Fisher, wrapped Cauchy, and elliptical families. In the narrow-capture limit of a vanishingly small target, directional persistence induces anomalous scaling, including regimes where the MFPT remains finite whereas standard diffusion would predict divergence. Finally, we obtain a Langevin representation that accurately reproduces first passage statistics. Analytical predictions are confirmed by numerical simulations.
The vertebrate limb provides an interesting system to study how tissue growth and molecular signaling interact to shape complex skeletal patterns. How these processes are coordinated across space and time is not fully understood. This study introduces a computational tool to examine how growth interacts with positional cues and self-organizing patterning mechanisms to shape skeletal structures in both mice and axolotl limbs. We developed the Growth-Reaction-Diffusion (GRD) framework, a reaction-diffusion system within a growing domain, where reaction represents the regulation of patterning cues and diffusion captures their spatial propagation. The relative contribution of growth, reaction and diffusion is modulated through two non-dimensional parameters, whose spatial variation is informed by positional cues derived from experimental morphogen maps. This formulation normalizes the reaction-diffusion equation relative to growth, enabling investigation of how different spatiotemporal regimes of growth interact with reaction and diffusion to produce whole limb patterning. The GRD framework captures the progressive formation of all limb segments: the humerus, radius/ulna, and the digits patterns. Our simulations indicate that in the proximal region (humerus, radius/ulna) the contributions of growth, reaction and diffusion are equally important to patterning, but in the distal elements (digits) the reaction and diffusion contributions are much greater than the contribution of growth to the formation of the digits. Through a single framework, we simulate the whole-limb skeletal patterns in both mice and axolotls, despite their morphological differences. These results highlight the model's potential to explore conserved and divergent features of limb development from an evolutionary perspective through a unified mechanism across species.
As urban areas expand in eastern USA, the convergence of historical and modern anthropogenic source inputs has resulted in a complex geochemical signature of road dust pollution, while representing a critical public health issue for communities. In this study, road dust collected at seven (7) cities in eastern USA was analyzed for 11 potential toxic elements (PTEs, e.g., Cu, Zn, As, Se, Ni, Fe, Mo, V, Co, Cd, Pb) and examined for their characteristics, sources, and potential health risks. Multivariate statistical analyses show the regional difference between northeastern (Trenton, NJ; Wilmington, DE), Piedmont (Richmond, VA; Raleigh, NC; Greensboro, NC), and southeastern cities (Charleston, SC; Augusta, GA; outlining the spatial variability of eastern USA  road dust sources. Above-unit enrichment factors (EFs > 1) from Cu, Zn, Mo, and Ni imply accumulation from non-natural sources, such as non-exhaust traffic emissions and industrial activities. Hazardous PTEs (e.g., Pb, As, Cd) exhibited EFs < 1, reflecting their historical input in surveyed cities, and were associated with low-income communities. Source apportionment approaches estimate a one-third contribution from hazardous PTEs (coal combustion, insecticide use) and two-thirds from other prominent urban sources (waste incinerators, vehicle emissions, and industrial activities). Trenton, Raleigh, and Greensboro also see a higher respirable dust fraction (< 10 µm) than other cities, leading to potentially higher inhalation health risk. Hazard index (HI) estimation shows overall 4.5-times higher values in children than adults across all cities, with Augusta, GA exhibiting elevated hazard exposure (HI > 1). Insights from this study revealed the spatial variability of road dust PTEs levels, complemented domestic legacy contaminant work, and revealed new source information for residential areas over the East Coast to highlight potential environmental impacts.
Time-varying coefficient modeling (TVCM), which represents regression coefficients as smooth functions of continuous time, provides a flexible framework for uncovering complex patterns of change in levels and associations in intensive longitudinal data. However, conventional TVCM remains limited to investigating directional effects across individuals. By introducing a TVCM formulation of the multivariate normal distribution, the present study extends TVCM to explore change in undirected associations (couplings) and variability, thereby broadening its utility for psychological research. We discuss three versions of this approach: an aggregate-level model and two hierarchical versions capturing interindividual differences in unfolding change, either via person-specific intercepts accounting for onset differences or through fully person-specific coefficient functions smoothed via partial pooling. To illustrate the proposed developments, we apply them to six weeks of intensive longitudinal data from 16 anxiety patients undergoing therapy and examine unfolding changes in the level and volatility of nervousness and threat monitoring, their coupling, as well as between-person heterogeneity in each of these. We further show how inspecting first-order derivatives of the coefficient functions supports identifying periods of stability and change. Finally, we discuss extensions incorporating person-level characteristics to explain heterogeneity in patterns of change and predict outcomes.
Variable thermal conductivity plays important role in precisely modeling of hybrid nanofluid (HNF) flow and heat transfer. In practical applications, thermal conductivity frequently varies with temperature, nanoparticle concentration, and fluid composition, affecting the rate of energy transfer within the system. The present investigation offers an inclusive investigation of heat transfer in magnetized flow of a HNF consisting of zinc (Zn) and silicon dioxide (SiO2) nanoparticles dispersed in water (H2O), towards a rotating stretchable surface. The management of thermal energy in rotating, high-speed environments can be significantly enhanced through the use of magnetized nanofluids with optimized nanoparticle shapes. This study incorporates the effects velocity slip, and convective boundary conditions. The principal objective is to explore the transport phenomena of heat and mass transfer, as well as the bioconvective behavior induced by motile microorganisms within the hybrid nanofluid. By applying similarity transformation, the governing flow model of PDE's for momentum, energy, concentration, and motile microorganism distribution are condensed to a set of coupled nonlinear ODE's. These equations are numerically solved with MATLAB software, which employs a vigorous shooting method. A thoroughly parametric investigation is performed to assess the influence of several physical parameters, including the magnetic field strength (Μ), rotational parameter (λ), velocity slip coefficient (β), thermal radiation parameter ([Formula: see text]), variable thermal conductivity (ε), and the volume fractions of the nanoparticles ([Formula: see text]), along with the nanoparticle shape factor. Furthermore, the effects of the Lewis number ([Formula: see text]), Peclet number ([Formula: see text]) and bioconvective Lewis number ([Formula: see text]) are examined with respect to species concentration and microorganism. The numerical fallouts are presented both graphically and in tabular format, demonstrating that enhancements in magnetic field strength and radiation parameter notably affect the thermal and flow characteristics. The legitimacy of the suggested model is confirmed through excellent agreement with benchmark outcomes from existing literature across various Prandtl number ([Formula: see text]) values, thereby signifying the reliability and practical relevance of the current approach in advanced hybrid nanofluid-based thermal management systems.
The evolution of a system coupled to baths is commonly described by a master equation that, in the long-time limit, yields a steady-state density matrix. However, when the same evolution is unraveled into quantum trajectories, it is possible to observe a transition in the scaling of entanglement within the system as the system-bath coupling increases-a phenomenon that is invisible in the trajectory-averaged reduced density matrix of the system. Here, we go beyond the paradigm of trajectories from master equations and explore whether a qualitatively analogous entanglement-scaling transition emerges in a single unitary evolution of the combined system-bath setup, without monitoring the dynamics of the system. We investigate the scaling of entanglement in a unitary quantum setup composed of a two-dimensional lattice of free fermions, where each site is coupled to a fermionic bath. As the system-bath coupling increases, the logarithmic fermionic negativity reveals an entanglement transition from logarithmic-law to area-law scaling. This occurs while the system's steady-state properties are trivial, highlighting that the signatures of these different scalings are within the bath-bath correlations. Evidence of the transition is also found in the mutual information and the correlations of the full system-bath setup, suggesting that the entanglement transition is underpinned by a change in the spatial structure of quantum information.
The impact of air temperature on COVID-19-related health outcomes, particularly in the context of extreme weather, remains underexplored. This time-stratified case-crossover study included 200,679 laboratory-confirmed COVID-19 emergency department (ED) visits in Ontario, Canada, from 2020 to 2023. Conditional logistic regression with distributed lag nonlinear models was used to assess the association of mean daily temperature with COVID-19 ED visits. Cumulative 0-5-day lag associations of cold and heat exposure were presented as odds ratios (ORs) with 95% confidence intervals (CIs) at the 1st and 99th percentiles of the temperature distribution, referenced to the minimum risk temperature (MRT). A non-linear, W-shaped relationship was observed between temperature and COVID-19 ED visits, with associations in extreme to very cold (-35.7°C to -21.6°C), cold (-14.1°C to -10.3°C), lightly cold to moderate (-10.2°C to 21.6°C), and warm to very hot (21.7°C to 29.2°C) temperature ranges (MRT: -17.2°C). Notably, heat exposure (24.6°C) showed stronger associations (OR: 1.53, 95% CI: 1.40-1.67) than cold exposure (-17.6°C, OR: 1.00, 95% CI: 0.996-1.005). Risk patterns varied across viral variants, with cold exposure associated with increased ED visits for Omicron and Wild variants, while heat exposure was linked to increased visits for Alpha, Delta, and Gamma variants. Associations were more pronounced for females, individuals aged <65 years, and those with a primary COVID-19 diagnosis. Short-term exposure to non-optimal temperatures significantly increases COVID-19-related emergency healthcare demand, with variant- and demographic-specific differences. Temperature-responsive public health strategies are needed to reduce the burden of COVID-19 and other infectious diseases as climate extremes intensify.
In times of epidemics, swift reaction is necessary to mitigate epidemic spreading. For this reaction, localized approaches have several advantages, limiting necessary resources and reducing the impact of interventions on a larger scale. However, training a separate machine learning (ML) model on a local scale is often not feasible due to limited available data. Centralizing the data is also challenging because of its high sensitivity and privacy constraints. In this study, we consider a localized strategy based on the German counties and communities managed by the related local health authorities (LHA). For the preservation of privacy to not oppose the availability of detailed situational data, we propose a privacy-preserving forecasting method that can assist public health experts and decision makers. ML methods with federated learning (FL) train a shared model without centralizing raw data. Considering the counties, communities or LHAs as clients and finding a balance between utility and privacy, we study a FL framework with client-level differential privacy (DP). We train a shared multilayer perceptron on sliding windows of recent case counts to forecast the number of cases in the future, while clients exchange only norm-clipped updates and the server aggregates updates with DP noise. We evaluate the approach on COVID-19 data on county-level during two phases: November 2020 and March 2022 (Omicron). As expected, very strict privacy ([Formula: see text]) yields unstable, unusable forecasts. At a moderately strong but still privacy-preserving level ([Formula: see text]), the DP model closely approaches the non-DP model: [Formula: see text] (vs. 0.96) and mean absolute percentage error (MAPE) [Formula: see text] in November 2020; [Formula: see text] (vs. 0.90) and MAPE [Formula: see text] in March 2022. Overall, our results support the feasibility of privacy-preserving collaboration among health authorities for local forecasting. In the evaluated COVID-19 phases, client-level DP-FL delivered useful county-level predictions with formal privacy guarantees under the stated threat model. The appropriate privacy budget should nevertheless be re-evaluated for other epidemic phases and applications.