With advancements in mobile sensing technologies, there is a growing need for scalable and interpretable stress monitoring solutions that remain robust over time. Existing smartphone passive sensing approaches rely on statistical app usage features or multi-modal sensor data, making them susceptible to distribution shift and feature evolution (e.g., schema drift), and adding model complexity. To address these challenges, we propose the Smartphone Human Interaction-based Routine Behavior Task Mining, Modeling, and Feature Extraction (SHIRBT-MMF) framework, which models stress-related behaviors by mining finegrained interaction routine tasks rather than aggregated app usage patterns. SHIRBT-MMF leverages multi-level sequential pattern mining and large language model-based automated task modeling to extract interpretable and stable features from within-app UI state transitions. Unlike traditional methods that require hundreds of apps, SHIRBT focuses on a small, consistent set of routine-based tasks, mitigating covariate shift and feature evolution while improving model robustness. We validated SHIRBT-MMF through one- and four-month in-the-wild datasets with 26 participants, demonstrating that the SHIRBT-based personalized model achieves an average accuracy of 75%, outperforming baseline models by 5% while using only 3-6% of app types. Additionally, SHIRBT features remain stable over time, reducing covariate shift and ensuring reliable performance. With its expandability to other mental health, interpretability, and privacy-conscious design, the SHIRBT-MMF framework lays the foundation for personalized digital mental health monitoring.
Real-time monitoring of methane can alert for explosion or health issues, but it is still difficult to detect methane with high sensitivity and low detection limit at room temperature. This work reports a highly sensitive room-temperature methane gas sensor based on heterojunction nanocomposites between n-type indium-zinc oxide (In2O3-ZnO) and p-type porous laser-induced graphene (LIG) prepared by laser direct writing and facile spray coating. The resulting In2O3-ZnO/LIG-based methane gas sensor exhibits a large response of -27.48% to 500 ppm methane room temperature, a fast response/recovery time of 48/169 s, and an experimentally demonstrated ultralow detection limit of 3 ppm. After confirming the stable response over 9 days and against bending with a radius of curvature down to 10 mm, the In2O3-ZnO/LIG gas sensors are further combined with a wireless powering and transmitting module into an integrated methane gas sensing device to wirelessly detect methane in real time for early warning and safe evacuation in mining and fertilizer plants.
Analysis of coding areas has long been used to study monogenic illnesses, but despite the extensive use of whole-exome sequencing (WES), up to half of suspected cases remain genetically unexplained. Variants outside coding areas can alter splicing, transcript stability, or gene regulation, compromising normal gene activity. These include mutations in noncoding RNAs, promoters, enhancers, deep intronic sequences, and untranslated regions (UTRs). Several well-known disorders have been linked to these mechanisms, including β-thalassemia caused by deep intronic mutations leading to aberrant splicing, familial hypercholesterolemia caused by promoter defects affecting LDLR expression, and inherited retinal diseases driven by noncoding variants influencing retinal gene regulation. These instances show that pathogenic variation is not limited to the exome and can have significant clinical implications. This review summarizes current understanding of noncoding and regulatory variants in monogenic diseases, discusses how they influence diagnosis and therapy, and highlights integrative approaches combining genomic, transcriptomic, and epigenomic data. Multi-layered research has increased diagnostic accuracy and unveiled new therapeutic potentials, although noncoding variations make the connection between genotype and phenotype more complex. Noncoding regions will need to be incorporated into standard diagnostic procedures to convert molecular insights into concrete therapeutic applications in the future. Predictive algorithms, patient-derived model systems, and functional validation testing will all help to simplify this process.
This study investigated the main causes of Air Force pilot fatigue, which could lead to aviation accidents. Data were collected from Korean pilots through in-person survey, utilizing the Human Factors Analysis and Classification System (HFACS). The study focused on preconditions for unsafe acts and unsafe supervision, as well as detecting fatigue using a modified checklist. A three-factor structure comprising 17 subfactors was identified, exhibiting robust psychometric properties, good model fit and reliability. Strong positive relationships were observed between pilot fatigue and all subfactors. The study highlighted the influence of age, job identity, and total flight hours on pilot fatigue. Instructors and more experienced pilots had higher fatigue scores than younger pilots. These findings underscore the need for targeted interventions to reduce fatigue among middle-aged pilots, instructors, and experienced pilots. The results validate the model's applicability in Korea and suggest its potential utility in other industrialized nations.
Understanding the sources of atmospheric particulate matter (PM) through reliable source apportionment methods is essential to support effective pollution control policies. Positive matrix factorization (PMF) has become the most widely used source apportionment method worldwide. However, conventional PMF analysis provides source profiles with uncertainties but contributions (g-value) without uncertainties limiting the utility of the resolved source-specific PM. This study estimates the g-value uncertainties in PMF-resolved PM2.5 source contributions by combining PMF with an effective variance least-squares (EVLS) analysis. Data from daily PM2.5 samples collected in Tianjin, China, were analyzed with the EPA PMF to identify the major sources. The EVLS method calculated the uncertainties in the g-value using the displacement intervals as the profile uncertainties. The primary PM2.5 contributors in Tianjin during 2013 to 2019 included secondary sulfate (25.39 +/- 7.51 mu g/m3), coal combustion (22.07 +/- 9.41 mu g/m3), secondary nitrate (19.88 +/- 4.74 mu g/m3), dust (16.12 +/- 7.36 mu g/m3), vehicle emissions (12.46 +/- 5.04 mu g/m3), the steel industry (12.32 +/- 8.58 mu g/m3), biomass burning (5.93 +/- 2.61 mu g/m3), and the galvanizing industry (3.37 +/- 2.89 mu g/m3). PMF-EVLS provided more comprehensive insights by better quantifying the contributions and uncertainties for each source, as shown by the more accurate reproduction of the original concentrations, particularly the higher concentration samples. Secondary sulfate and coal combustion had the highest uncertainties. This three-step method presents a significant advancement in PM source apportionment by allowing contribution uncertainty estimation, thus enhancing the reliability of source apportionment results and supporting data-driven air quality management decisions. Conventional PMF source apportionments rarely quantify the uncertainties in source contributions. This study integrates effective variance least squares with PMF to provide robust error estimates for each source-specific contribution, addressing this critical gap and enhancing the reliability of results for air quality management.