Systemic BCG administration induces transcription factor signature in CD4+ T cells that cooperates with IL-12 signaling to drive antiviral Th1 differentiation.

Online geocoding platforms are essential tools for location-based services (LBS) and spatial analysis because they convert textual addresses into geographic coordinates. However, their results often exhibit inconsistencies due to variations in geospatial databases and geocoding algorithms across platforms. A key challenge is the lack of standardized criteria to evaluate and integrate multi-source geocoding outputs. Integrating multi-platform results with public road network data may offer a viable path toward higher accuracy. Building on this idea, this study proposes a road-constrained optimization method for web geocoding. It utilizes results from three popular platforms (Baidu, Amap, and Tencent) and employs open-source road network data to impose spatial constraints. A rigorous spatial matching and filtering process is designed to eliminate erroneous coordinates and optimize results. We tested 3,000 addresses in Nanjing, China, comparing errors before and after applying our method. Results show the algorithm reduces errors from 112.36–210.58m to 72.07m on average. Error distribution also becomes more tightly clustered around the ground-truth locations. Statistical analysis further reveals that errors exceeding 500m can be corrected to within 100m. This demonstrates a significant accuracy improvement and offers a practical solution for users of online geocoding services.

Exploring factors influencing open government data value realization in China: A mixed design using grounded theory, system dynamics, and questionnaire survey

ABSTRACT This paper examines whether public data openness affects insider trading behaviour and profitability. Utilizing China’s staggered municipal launch of open government data (OGD) platforms and a stacked difference-in-differences design, we find that public data openness reduces insider trading profitability. Mechanism tests reveal that OGD platforms operate through an information channel by enriching the firm-level information environment, specifically increasing analyst coverage, narrowing bid-ask spreads, and reducing stock price synchronicity. Further analyses show that OGD platforms compress insider trading on both the extensive margin (reducing overall trading likelihood) and the intensive margin (lowering the probability of trade success conditional on trading). Cross-sectional tests show that this deterrent effect is more pronounced for firms with weaker corporate governance and those located in regions with superior digital infrastructure. These findings offer insights for policymakers that public data infrastructure generates positive externalities for financial market integrity.

Reference MS/MS libraries remain incomplete due to the vast chemical diversity of metabolites, leaving many spectra from untargeted metabolomics experiments unannotated─the "dark matter" of metabolomics. Machine learning can extend metabolite annotation beyond direct library matches, but its success depends critically on how MS/MS spectra are converted into numerical representations that capture chemically meaningful features while reducing sparsity. Although numerous spectral representations exist, they have not been systematically compared. Using over 71,000 unique compounds with merged-energy MS/MS spectra, we benchmarked a broad set of spectral featurization methods, including fixed and adaptive binning, global-quantile variable-width bins, frequent-peaks representations, spectrum hashing, and learned embeddings such as Spec2Vec, MS2DeepScore, DreaMS, and SpecEmbedding. We further evaluated how vector dimensionality affects performance. A total of 105 neural network models were trained under 5-fold cross-validation to predict Mol2Vec molecular embeddings and retrieve correct structures from a 0.6-million-compound database. Retrieval was assessed at 0.1, 3, and 10 ppm mass tolerances, and a null ranking model was generated to determine expected Top-N accuracy under random candidate ordering. Adaptive binning, frequent-peaks, and DreaMS produced the most accurate embedding predictions. On the test data set, Top-1 retrieval reached 46%, 44%, and 38% for 0.1, 3, and 10 ppm, respectively, with Top-5 accuracies up to 77%. In the CASMI2022 data set, Top-1 performance remained similar at 0.1 ppm but dropped markedly at wider tolerances, reaching only 26% at 3 ppm and 23% at 10 ppm. To ensure reproducibility and broad community applicability, results were further validated on two fully open benchmark data sets, MassSpecGym and Spectraverse, with findings consistent across all three resources. These results underscore clear performance differences among featurization strategies, the strong dependence of retrieval accuracy on mass precision, and the need for evaluation metrics aligned with structure-level annotation tasks.

Passive location tracking through smartphone-based services offers promising avenues for mobility research, yet its reliability and analytical value remain underexplored. This study presents a comprehensive methodology for extracting commuting patterns from Google Maps Timeline (GMT) data and applies it to a 30-day longitudinal dataset on 294 participants, collected in 2024 in Southern Switzerland. We developed and released an open-source data processing methodology to convert raw GMT activity segments into discrete location bins, enabling the detection of significant places (home and workplace) and commuting trips through unsupervised algorithms. Our findings show that GMT achieves a mean location coverage of 72% across study participants and days with a 200-meter distance threshold between consecutive segments, and allows for the reliable identification of commuting trips in days in which study participants self-reported having worked. While GMT presents limitations in data completeness and algorithm transparency, it offers scalable, low-burden alternatives to traditional mobility data collection methods. Future research should explore its applicability across diverse populations and contexts, and consider open-source alternatives for enhanced transparency and control. • Google Maps Timeline (GMT) provides a passive solution for location tracking. • We provide methods to process raw GMT data into significant places and commuting trips. • We test our methods on real-world data from 294 people in Southern Switzerland. • We find a GMT coverage of 72%, and reliable home and workplaces for 65% of study participants. • GMT is a low-burden tool complementing self-reports for policy making and large-scale mobility research.

Recent advancements in machine learning-based signal analysis, coupled with open data initiatives, have fuelled efforts in automatic sleep stage classification. Despite the proliferation of classification models, few have prioritised reducing model complexity, which is a crucial factor for practical applications. In this work, we introduce Multi-Scale and Attention Convolutional Neural Network (MSA-CNN), a lightweight architecture featuring as few as ∼ 10,000 parameters. MSA-CNN leverages a novel multi-scale module employing complementary pooling to eliminate redundant filter parameters and dense convolutions. Model complexity is further reduced by separating temporal and spatial feature extraction and using cost-effective global spatial convolutions. This separation of tasks not only reduces model complexity but also mirrors the approach used by human experts in sleep stage scoring. We evaluated both small and large configurations of MSA-CNN against nine state-of-the-art baseline models across three public datasets, treating univariate and multivariate models separately. Our evaluation, based on repeated cross-validation and re-evaluation of all baseline models, demonstrated that the large MSA-CNN outperformed all baseline models on all three datasets in terms of accuracy and Cohen’s kappa, despite its significantly reduced parameter count. Lastly, we explored various model variants and conducted an in-depth analysis of the key modules and techniques, providing deeper insights into the underlying mechanisms. The code for our models, baselines, and evaluation procedures is available at https://github.com/sgoerttler/MSA-CNN . • This study develops a lightweight and interpretable CNN for sleep stage classification. • Multi-scale temporal convolutions are introduced to extract spectral features with fewer parameters. • We demonstrate that attention improves performance and captures known sleep pattern interactions.

The human microbiome is a dynamic ecosystem that profoundly influences host physiology through complex molecular interactions. Advances in high-throughput profiling now enable multi-omics measurements at scale, yet integration remains difficult due to biological complexity, technical variability, sparsity, and small cohorts. This review targets bioinformatics practitioners and clinical microbiology researchers applying machine learning to host-microbiome studies. Here, we survey state-of-the-art methods for integrating heterogeneous data types and highlight algorithmic innovations for high dimensionality and small cohorts. We also examine approaches for interpretability that translate mechanistic insight into clinically actionable models. Finally, we outline a standardized benchmarking framework emphasizing open data, rigorous evaluation, and biologically informed architectures. By synthesizing multi-omics measurements with advanced analytics, we chart a pathway toward personalized, microbiome-based therapies while deepening our understanding of host-microbiome crosstalk.

Mental health conditions in Mexico: Diagnosis trends based on hospital records.

Machine learning early warning for urban heat risk with CMIP6 projections.

Prior evidence suggests that journals requiring open data are associated with higher levels of data sharing in the published psychology literature. Data sharing policies are not, however, consistently implemented or enforced. The American Psychological Association (APA), in 2020, signed onto the Transparency and Openness Promotion Guidelines, which promote increasingly stringent data sharing policies. The current study examined self-reported data sharing in APA journals and whether stricter policies are linked to higher levels of self-reported data sharing. We assessed self-reported data sharing practices in 1,250 articles published between 2023 and 2025 in 25 APA journals. Using logistic regression, we examined the association between journal policy level and self-reported data sharing. We then applied post-stratification weighting, based on the actual distribution of policy levels across APA journals and their associated percentages of data sharing, to estimate the overall percentage of self-reported data sharing. We estimated overall self-reported data sharing to be 30.3% (95% CI [27.7, 33.0]). Journal policy stringency was strongly associated with self-reported data sharing: among journals with no data sharing policy, 15.0% of articles reported shared data (30 out of 200; 95% CI [10.7, 20.6]); among journals that mandate that authors reveal whether they shared their data, 26.4% of articles reported shared data (145 out of 550; 95% CI [22.8, 30.2]), and among journals that mandate data sharing, 70.4% of articles reported shared data (352 out of 500; 95% CI [66.2, 74.2]). At the same time, even journals with more stringent data sharing requirements did not consistently enforce their policies. These results indicate that despite widespread endorsement of open science practices, data sharing is not yet a customary practice among psychology researchers. The observed self-reported data sharing practices do, however, show that more stringent journal policy requirements are associated with increased openness. While the observational nature of the present study precludes strong causal inferences, stricter journal policies may represent one potential factor relating to the adoption of data sharing practices, including the use of data embargoes and greater transparency around decisions not to share data. Overall, findings highlight a promising direction for ongoing efforts to promote openness in data sharing, while underscoring the need for future research to clarify the mechanisms that link policy to practice.

Supervised deep learning (DL) receives great interest for automated analysis of microscopic images with an increasing body of literature supporting its potential. The development and testing of those DL models rely heavily on the availability of high-quality, large-scale data sets. However, creating such data sets is a complex and resource-intensive process, often hindered by challenges such as time constraints, domain variability, and risks of bias in image collection and label creation. This review provides a comprehensive guide to the critical steps in data set creation, including (1) image acquisition, (2) selection of annotation software, and (3) annotation creation. For image acquisition, besides ensuring a sufficiently large number, it is important to address sources of image variability (domain shifts), such as those related to slide preparation and digitization, that could lead to algorithmic errors if not adequately represented in the training data. For annotations, key quality criteria are the 3 "C"s: correctness, consistency, and completeness. For mitigation of annotation bias of a single annotator, this review explores advanced annotation methods (eg, computer-assisted annotations). To support data set creators, a standard operating procedure is provided as supplemental material, summarizing all important considerations for data set creation. Furthermore, this article underscores the importance of open data sets in driving innovation and enhancing reproducibility of DL research. By addressing the challenges and offering practical recommendations, this review aims to advance the creation and availability of high-quality, large-scale data sets, ultimately contributing to the development of generalizable and robust DL models for pathology applications.

Cannabis cue reactivity paradigms are instrumental in studying the behavioral and neurocognitive mechanisms of cannabis use and cannabis use disorders; however, image sets used for cannabis cue reactivity paradigms vary between studies, and the lack of reliability and validity assessment hinders the quality of evidence they generate. The main aim of this study was to create a novel, open access, standardized and representative database of cannabis use-related images including control images matched by resolution, luminosity and complexity: The Cannabis Research Image Database (CRESIDA). The secondary aim was to examine whether subjective cannabis cue-induced craving was associated with cannabis use severity and whether this relationship was moderated by image type. As an illustrative example of how our open data can be used and how sample characteristics can shape cue reactivity, we also explored the role of cannabis-tobacco mixing by comparing cannabis cue induced cannabis and tobacco craving between individuals who did and did not mix the substances. An online survey was administered to participants recruited via online platforms, community advertisement and snowballing. USA, the Netherlands and Australia. 689 participants who consumed cannabis monthly to daily (385 men, 298 women, 6 other) were recruited between January 2022 and May 2024. Out of 93 cannabis images and 93 matched neutral images, participants each rated 31 image pairs for cannabis craving (the primary outcome), arousal, valence and tobacco craving. Participants were characterized for socio-demographic data, level of cannabis use and related problems and mixing cannabis and tobacco. A subset of 78 images was selected for further analysis based on cannabis craving results. Image ratings were evaluated for internal consistency (α). Furthermore, we examined the association between cannabis cravings and cannabis use characteristics, and explored if cannabis craving ratings were affected by image type (i.e. product, paraphernalia and actions) and by using cannabis alone vs. mixing cannabis and tobacco. The database showed excellent reliability (α=0.995-0.965). Cannabis craving, valence and arousal discriminated cannabis and control images. More cannabis use days [unstandardized beta (β) =0.162, P < 0.001] and cannabis use-related problems (β =0.268, P < 0.001) were statistically significantly associated with higher image-related cannabis craving. Mixing cannabis with tobacco, compared with using cannabis alone, was associated with the presence of tobacco craving in relation to cannabis images, and with greater cannabis craving in relation to cannabis images (β = -0.457, P < 0.001). Images in the open access Cannabis Research Image Database (CRESIDA, https://osf.io/dc9nz/) appear to be reliable and valid for the scientific study of cue reactivity internationally, providing a broad range of free to use cannabis and control images.

ABSTRACT This article reconceptualizes environmental data activism as situated practices that mediate actions through data infrastructures and translation. Drawing on participant observation, interviews, and document analysis of two environmental NGOs (ENGOs) in Taiwan and their collaborations with civic-tech communities, we identify three key roles of ENGOs in environmental data activism: (1) extracting and infrastructuring datasets; (2) translating and re-embedding data into everyday practices through digital tools such as mobile apps and crowdsourcing games; (3) and mediating between different social actors in data activism. Hence, we theorize ENGOs as civic data intermediaries that turn open data into civic data infrastructures. We further identify contingent civic alliances in data activism. These are volunteer-driven collaborations that are generative yet precarious, producing episodic surges of both innovation and breakdown. This article argues that the value of data emerges not from purity of measurement but from socio-technical work that renders data publicly usable and criticizable. Conceptually, the research advances a practice-based theory of data justice as infrastructural work, showing how civic groups convert data from records into connected actions in everyday activism. From this perspective, the limits of justice are set as much by the labor of maintenance, temporal rhythms, and alliance fragility as by the epistemic accuracy of data.

ABSTRACT The concept of digital sovereignty is a key framework for analysing geopolitical, economic and ideological disputes over technological infrastructures in contemporary capitalism. From a critical perspective, digital sovereignty in education is a contested field where state, corporate and social-movement actors struggle to define how digital infrastructures should be governed, owned and operationalised. This article analyses two configurations of digital sovereignty in Spanish education policy: ALIA, a state-led public AI infrastructure within Spain’s national AI strategy, and Democratic Digitalisation (DD), a social-movement-driven platform alternative to Google’s dominance. Methodologically, the study combines Critical Discourse Analysis, Policy Network Analysis and Technographic Analysis to examine how digital sovereignty is articulated through policy discourse, governance relations and socio-technical configurations. Findings show that ALIA translates public autonomy into an AI project, but its operationalisation remains entangled with market-oriented agendas and public-private governance that limit state-led sovereignty. Meanwhile, DD advances technological sovereignty through open-source software, auditability, interoperability and community data control. However, through the lens of critical digital sovereignty in education, its transformative potential remains limited insofar as infrastructural autonomy is not fully connected to teachers’ labour, curricular democracy and collective participation in school governance.

Ficus carica L. is an economically important fruit crop widely cultivated in the Mediterranean region. In this study, genetic diversity and population structure were investigated in 72 F. carica genotypes collected from the Derecik and Çukurca regions of Hakkâri province, Türkiye, using 15 highly polymorphic Start Codon Targeted (SCoT) markers. A total of 481 amplification bands were obtained, of which 475 were polymorphic, resulting in a high average polymorphism rate of 98.60%. Genetic diversity indices indicated substantial variation among the genotypes, with a mean effective number of alleles of 1.53, gene diversity of 0.31, and Shannon information index of 0.47. The average genetic distance among genotypes was 0.37, with the highest pairwise distance (0.721) observed between genotypes HC4 and HD7. Analysis of molecular variance (AMOVA) revealed that most genetic variation was distributed within populations (93%), whereas only 7% was among populations. Bayesian STRUCTURE analysis identified two distinct genetic clusters corresponding largely to geographic origin, with 22 genotypes (30.56%) classified as admixed based on a membership coefficient threshold of < 0.70. Principal coordinate analysis (PcoA) clearly separated genotypes according to their sampling locations, where Axis 1 and Axis 2 explained 24.31% and 15.07% of the total genetic variation, respectively. Overall, these findings demonstrate the effectiveness of SCoT markers in assessing genetic diversity and population structure in F. carica germplasm from southeastern Türkiye. Future studies should use codominant markers (SSRs and SNPs) expand geographic sampling, and adopt open data repositories to enhance conservation and breeding strategies.

The United Kingdom (UK) and Dutch governments have recently implemented mandatory financial risk (affordability) assessments for online gambling as a harm prevention measure. Assessments should trigger at a level of gambling expenditure that strikes a balance between harm prevention (most at-risk consumers should surpass the threshold) and liberty preservation (most no-/lower-risk consumers should gamble below it), yet little empirical research exists to guide threshold setting. We aimed to demonstrate how research can inform the harm-prevention, liberty-preservation trade-off in this context and evaluate the UK's proposed implementation of financial risk assessments. We reanalysed a dataset that combines self-reported Problem Gambling Severity Index (PGSI) scores and open banking data from consumers who gamble (n = 424) to (1) simulate the impact of the UK's rolling 30-day £150 net-deposit (deposits minus withdrawals) threshold for financial risk assessments, and (2) identify optimal threshold values for these assessments under different circumstances. Participants were UK residents who had gambled in the past year, recruited via Prolific in April 2024. Participants completed a survey containing the PGSI and agreed to provide their past 12 months' banking records. Over 12 months, two-thirds of at-risk (PGSI ≥1) and nearly half of no-/lower-risk participants crossed the UK's £150 threshold [area under the curve = 0.66, 95% confidence intervals (CIs) = 0.62-0.71], demonstrating a greater emphasis on harm prevention over liberty preservation. Increasing the value to £186.9 (95% CIs = £69.5-£401.7) slightly improved this balance, although £150 remained within the range of appropriate values. Optimising for harm prevention in our sample required lowering the threshold to £39.0 (95% CIs = £29.6-£58.8), while emphasising liberty preservation increased it to £716.5 (95% CIs = £508.5-£990.9). We found that using a more conservative definition of risk (≥2 PGSI harms) resulted in higher thresholds, and lower thresholds may be appropriate for younger adults (<30 years). Finally, our findings suggest that thresholds based on spending with all operators-rather than single operators as implemented in the UK-may be better able to differentiate at-risk from no-/lower-risk consumers, although the added benefit of this approach in our sample was marginal and further research is needed to confirm its value. The United Kingdom's £150 net-deposit threshold for financial risk assessments for online gambling may place more emphasis on harm prevention than liberty preservation. This study provides a methodological template for guiding the implementation of financial risk assessments for online gambling. Because our sample is not representative of the broader UK gambling population, our specific threshold estimates should be treated as provisional.

In October 2024, Parties to the United Nations Convention on Biological Diversity agreed to a new multilateral mechanism to fund biodiversity conservation through the sharing of benefits from open biodiversity data. Biological databases hosting genetic and other biological data, known as digital sequence information (DSI), are central to the implementation of the mechanism. This paper assesses the new international agreement and its implications for DSI databases. We walk through the database provisions in COP16 Decision 16/2, which include notifying users and submitters about the mechanism, improving metadata on geographical location of sample collection, and consistency with open access, as well as consideration of the FAIR, CARE, and TRUST principles. Drawing on surveys, interviews, and a workshop with biological database managers, we identify practical and scalable measures including updating terms of use, revising submission procedures, and strengthening user communication. We also propose approaches to capture and report non-monetary benefits such as capacity building, publications, interoperability, and training. These actions illustrate how DSI databases can remain open, sustainable, and globally connected while supporting benefit-sharing from the use of DSI on genetic resources.

This article examines whether, in the period 2015–2024, Poland was narrowing the gap separating it from the European Union in the areas of human capital, digitalisation, and the potential for artificial intelligence uptake. The empirical basis comprises ten indicators grouped into four domains - human capital, digital infrastructure and maturity, innovation potential, and economic outcomes - drawn from the World Bank Open Data API, OECD databases, and OECD.AI, each juxtaposing Poland with the Union aggregate. Descriptive statistics are employed: terminal values, the absolute gap, a catching-up index (the ratio of the national to the Union value), and the compound annual growth rate, deliberately forgoing econometric modelling. The results indicate pronounced yet heterogeneous convergence: Poland has steadily narrowed the gap in the domain of enabling conditions - research personnel, R&amp;D expenditure, basic digital infrastructure and the digital maturity of enterprises, and labour productivity - while diverging in the adoption of frontier technologies, including the direct use of artificial intelligence in firms. The findings are interpreted in the light of the general-purpose technology concept and the absorptive capacity of the economy.

High-resolution mass spectrometry (HRMS) is the gold-standard technique for comprehensively profiling chemical exposures in complex human matrices, making it a powerful analytical tool for advancing human exposome research. Yet the scarcity of HRMS reference data, including collision cross-section (CCS) measurements from ion mobility-mass spectrometry (IM-MS) and MS/MS fragmentation spectra, hinders confident structural annotation of chemical exposure agents across laboratories. We therefore developed ToxBase, a multidimensional (m/z, retention time, CCS, MS/MS) reference database for over 2,000 chemicals sourced from the U.S. Environmental Protection Agency's ToxCast chemical library. Built via high-throughput liquid chromatography-ion mobility-tandem mass spectrometry (LC-IM-MS/MS), the ToxBase database comprises 3,598 precursor ions spanning 2,075 unique compounds with excellent precision (98.5% of compounds display interday CCS RSDs < 1%) and strong cross-platform agreement. A high-quality MS/MS reference library of the fragmented precursors was assembled using targeted data-dependent acquisition and DDARawProcessor, a novel data extraction algorithm. When applied to LC-IM-MS/MS data obtained from human plasma, urine, and fecal samples (n = 20 per matrix), ToxBase rapidly enabled 42 high-confidence (Level 1) identifications. The ToxBase database is freely available and compatible with the open-source MS data processing platform Skyline for vendor-agnostic suspect screening workflows, providing a valuable resource for standardized, large-scale exposome analysis.