University of California, Berkeley

Background Depression and anxiety disorders are highly prevalent among young adults, with evidence suggesting sleep problems as key risk factors. Objective This study aimed to examine the association between insomnia and sleep characteristics with major depressive episode (MDE) and anxiety disorders, and the association after accounting for baseline mental health symptoms. Methods We conducted a prospective cohort study using data from the Students’ Health and Wellbeing Study (SHoT), surveying Norwegian higher education students aged 18 to 35 (N = 53,362). A diagnostic assessment of 10,460 participants was conducted in 2023. Self-reported insomnia, sleep duration, sleep onset latency, and wake after sleep onset were recorded in 2022. MDE and five types of anxiety disorders were assessed after one year using a self-administered CIDI 5.0. Analyses adjusted for age, sex, baseline mental health symptoms, and somatic conditions. Results Insomnia in young adults was associated with a significantly increased risk of MDE (adjusted RR = 3.50, 95 % CI = 3.18–3.84) and generalized anxiety disorder (GAD) (adjusted RR = 2.82, 95 % CI = 2.55–3.12) one year later. Sleep duration showed a reversed J-shaped association with mental disorders, with both short and, to a lesser extent, long sleep durations linked to elevated risks, even after adjusting for baseline mental health symptoms and somatic conditions. Although the associations were attenuated after adjustment, they remained statistically significant. Conclusion Sleep disturbances, including insomnia and abnormal sleep durations, predict mental health issues in young adults, even after accounting for baseline mental health and somatic health. Addressing sleep problems early may help prevent subsequent mental health conditions in this population.

Introduction: U.S. maternal mortality rates are two to three times higher than other high-income countries, with most deaths occurring postpartum. Fragmented care, exacerbated by health insurance gaps and workforce shortages, underscores systemic deficiencies. Although patients’ and clinicians’ perspectives are well-documented, little is known about payors’ and purchasers’ perspectives. Given their influence in coverage decisions, payment rates, and service reimbursement, the objective was to explore their perspectives and identify challenges and opportunities in improving postpartum care in California, currently engaged in reshaping maternal health pathways. Methods: We conducted a qualitative study using semi-structured interviews with high-level administrators from major California health insurance providers and purchasers between June and October 2023. Participants, recruited through professional connections, were selected through purposive sampling based on their involvement in maternal and child health coverage decisions. A hybrid inductive–deductive approach was employed to identify major themes. Results: Participants (n = 11) identified barriers including limited insurance coverage, lack of clinical provider incentives, reimbursement concerns, and misaligned measures and metrics. Opportunities to improve postpartum care focused on visit timing and frequency, alternative payment models, and improving continuity of care between birth and the transition to primary care. Conclusions: Insurance payors and purchasers identified postpartum care barriers and suggested solutions well-supported by the literature. These solutions—including reimagining global bundle payment models, updating Healthcare Effectiveness Data and Information Set measures, and promoting dyadic models—could address barriers, improve outcomes, and inform California’s ongoing maternal health transformation and those happening around the United States.

We study the problem of maximizing information divergence from a new perspective using logarithmic Voronoi polytopes. We show that for linear models, the maximum is always achieved at the boundary of the probability simplex. For toric models, we present an algorithm that combines the combinatorics of the chamber complex with numerical algebraic geometry. We pay special attention to reducible models and models of maximum likelihood degree one.

Artificial intelligence (AI) has emerged as a transformative force in the art world, fostering new avenues for creativity and challenging traditional notions of artistic authenticity. In 2024 February, the release of Sora by OpenAI symbolized a major breakthrough in the quality of AI-Generated Content (AIGC), calling for new study of up-to-date technique advancement. The aim of this study is to examine how public perceptions of AI art have evolved in the context of Sora's release and to identify the possible contributing factors, such as the release of Sora and media framing. A mixed-methods approach, using both qualitative tools (MDCOR and SENA) and qualitative analysis (digital ethnography), was employed to draw insights from YouTube comments. Findings indicate that Soras release contributes to a broader range of public opinion and a trend of positive sentiment. Topics discussed by the public shift from the use of AI in the artistic creation process to the multidimensional applications of AI art products across diverse industries and their consequent social impacts. The majority of people hold a more positive attitude, particularly in terms of trust, towards AI art, while some artists express fear of job replacement and ethical issues. Media framing of political preference and professional interests also plays an important role in shaping public opinion of AI art. This study carries several implications for policymakers, artists, and technologists, and advocates legislative protection for human creators while fostering innovation.

Importance State abortion policies may influence the practice locations of obstetricians and gynecologists (OBGYNs), having potentially significant implications for access to and quality of reproductive health care. Objective To explore changes in OBGYN practice locations from before to after the Dobbs v Jackson Women’s Health Organization US Supreme Court decision in June 2022. Design, Setting, and Participants National Plan & Provider Enumeration System data files were used in a descriptive cohort study assessing the association between state abortion policy environments and OBGYN practice locations in the US from January 1, 2018, to September 30, 2024, for all OBGYNs listed in the data files during the study period. Main Outcome and Measures The number of OBGYNs practicing in states with differing abortion laws and the movement of OBGYNs between these states before and after the Dobbs decision. Results The sample included 60 085 OBGYNs (59.7% women), of whom 3.8% were maternal-fetal medicine specialists and 12.9% were recent residency graduates. The mean increase in the per-quarter number of OBGYNs from before to after Dobbs was 8.3% (95% CI, 6.6%-10.1%) in states with total abortion bans, 10.5% (95% CI, 8.1%-13.0%) in states with gestational age limits or threatened bans, and 7.7% (95% CI, 5.9%-9.4%) in states with abortion protections. From the quarter immediately before Dobbs to the end of the study period, 95.8% of OBGYNs remained in protected states, 94.8% (95% CI, 94.3%-95.2%) remained in states threatening bans, and 94.2% (95% CI, 93.7%-94.7%) remained in states with abortion bans. Conclusions and Relevance In this descriptive cohort study, there were no significant differences in trends in OBGYNs’ practice locations across states with different abortion-related policy environments after the Dobbs decision. Although these findings do not provide insight into changes in the quality of care provided, they suggest that there are no major changes in the supply of OBGYNs associated with the Dobbs decision.

We exploit analogies between first-order algorithms for constrained optimization and non-smooth dynamical systems to design a new class of accelerated first-order algorithms for constrained optimization. Unlike Frank–Wolfe or projected gradients, these algorithms avoid optimization over the entire feasible set at each iteration. We prove convergence to stationary points even in a nonconvex setting and we derive accelerated rates for the convex setting both in continuous time, as well as in discrete time. An important property of these algorithms is that constraints are expressed in terms of velocities instead of positions, which naturally leads to sparse, local and convex approximations of the feasible set (even if the feasible set is nonconvex). Thus, the complexity tends to grow mildly in the number of decision variables and in the number of constraints, which makes the algorithms suitable for machine learning applications. We apply our algorithms to a compressed sensing and a sparse regression problem, showing that we can treat nonconvex $\ell ^p$ ℓ p constraints ( $p<1$ p < 1 ) efficiently, while recovering state-of-the-art performance for p=1 p = 1 .

This study explores the role of generative AI (GenAI) in providing formative feedback in children's digital learning experiences, specifically in the context of mathematics education. Using multimodal data, the research compares AI‐generated feedback with feedback from human instructors, focusing on its impact on children's learning outcomes. Children engaged with a digital body‐scale number line to learn addition and subtraction of positive and negative integers through embodied interaction. The study followed a between‐group design, with one group receiving feedback from a human instructor and the other from GenAI. Eye‐tracking data and system logs were used to evaluate student's information processing behaviour and cognitive load. The results revealed that while task‐based performance did not differ significantly between conditions, the GenAI feedback condition demonstrated lower cognitive load and students show different visual information processing strategies among the two conditions. The findings provide empirical support for the potential of GenAI to complement traditional teaching by providing structured and adaptive feedback that supports efficient learning. The study underscores the importance of hybrid intelligence approaches that integrate human and AI feedback to enhance learning through synergistic feedback. This research offers valuable insights for educators, developers and researchers aiming to design hybrid AI‐human educational environments that promote effective learning outcomes. Practitioner notes What is already known about this topic? Embodied learning approaches have been shown to facilitate deeper cognitive processing by engaging students physically with learning materials, which is especially beneficial in abstract subjects like mathematics. GenAI has the potential to enhance educational experiences through personalized feedback, making it crucial for fostering student understanding and engagement. Previous research indicates that hybrid intelligence that combines AI with human instructors can contribute to improved educational outcomes. What this paper adds? This study empirically examines the effectiveness of GenAI‐generated feedback when compared to human instructor feedback in the context of a multisensory environment (MSE) for math learning. Findings from system logs and eye‐tracking analysis reveal that GenAI feedback can support learning effectively, particularly in helping students manage their cognitive load. The research uncovers that GenAI and teacher feedback lead to different information processing strategies. These findings provide actionable insights into how feedback modality influences cognitive engagement. Implications for practice and/or policy The integration of GenAI into educational settings presents an opportunity to enhance traditional teaching methods, enabling an adaptive learning environment that leverages the strengths of both AI and human feedback. Future educational practices should explore hybrid models that incorporate both AI and human feedback to create inclusive and effective learning experiences, adapting to the diverse needs of learners. Policymakers should establish guidelines and frameworks to facilitate the ethical and equitable adoption of GenAI technologies for learning. This includes addressing issues of trust, transparency and accessibility to ensure that GenAI systems are effectively supporting, rather than replacing, human instructors.

People routinely choose between options varying on multiple attributes – homes to rent, movies to watch, and so on. Here, we test how much awareness people have of the mental processes underlying these choices. We develop a method to quantify awareness of value-based multi-attribute choice processes that accounts for diverse choice strategies. Across five studies, participants make choices and then report how they believe they made them. We use computational modeling to identify the process revealed in their choices, and compare it to their self-reports to quantify individuals’ accuracy about their choice process. While we observe substantial variation in accuracy, participants are often highly accurate about their choice process – more accurate than predicted by a sample of decision scientists – and more accurate than informed third-party observers, suggesting evidence for introspection. These results challenge notions that we are strangers to ourselves and instead suggest that people often know how they made value-based choices.

This paper draws on frameworks from the philosophical study of epistemic injustice and oppression to explore the epistemic manifestations of carcerality. We argue that people with histories of involvement with the carceral state (system-involved people) experience a distinctive array of epistemic exclusions that amount to epistemic oppression, and that this oppression is one mechanism by which the carceral state sustains and perpetuates itself. We introduce the term epistemic carcerality to refer to this form of oppression that is endemic to the carceral state. Using methods of empirically-engaged philosophy, we explore the contours of epistemic carcerality in the context of higher education both within carceral institutions (e.g., prisons) and on college campuses. We aim to establish epistemic carcerality as a valuable concept that identifies significant and unique epistemic harms encountered and resisted by system-involved people, and to establish it as a pressing concern for higher education institutions.

Root gravitropism relies on gravity perception by the root cap and requires tightly regulated phytohormone signaling. Here, we isolate a rice mutant that displays root coiling in hydroponics but normal gravitropic growth in soil. We identify COILING ROOT IN WATER 1 (CRW1) encoding an ETHYLENE-INSENSITIVE3 (EIN3)-BINDING F-BOX PROTEIN (OsEBF1) as the causative gene for the mutant phenotype. We show that the OsCRW1-EIN3 LIKE 1 and 2 (OsEIL1/2)-ETHYLENE RESPONSE FACTOR 82 (OsERF82) module controls the production of reactive oxygen species in the root tip, subsequently impacting root cap stability, polar localization of PIN-FORMED 2 (OsPIN2), symmetric distribution of auxin, and ultimately gravitropic growth of roots. The OsEIL1/2-OsERF82 ethylene signaling module is effectively impeded by applying gentle mechanical resistance to root tips, including growing in water-saturated paddy soil. We further show that mechanosensing-induced calcium signaling is required and sufficient for antagonizing the ethylene signaling pathway. This study has revealed previously unanticipated interplay among ethylene, auxin, and mechanosensing in the control of plant gravitropism.

Peers are individuals with lived experience of mental health challenges trained to provide support to others with similar challenges. Help@Hand was a multi-site project that integrated peers into digital mental health intervention (DMHI) implementation. This study uses the Consolidated Framework for Implementation Research (CFIR) to frame challenges reported by peers when implementing DMHIs. Individuals leading the local peer workforce completed quarterly online surveys about perceived challenges to DMHI implementation. Biannual interviews probed for details on survey-reported challenges. 103 quarterly surveys and 39 bi-annual interviews were collected from key informants at 11 Help@Hand sites between Summer 2020 and Fall 2022. One challenge was tied directly to DMHIs; namely, device distribution. Several related to the Implementation Process, including challenges with recruiting qualified peers and integrating peers into DMHI implementations; communication and collaboration; and translation. Challenges in the Individual domain included unclear peer roles and multi-tasking across various projects. Inner Setting challenges included structural barriers to hiring peers, issues with communication and project management, and workforce turnover. Outer Setting challenges related to environmental technology readiness, COVID-19, unclear decision-making processes across the collaborative, and uneven communication between sites’ peers. Funding uncertainty bridged the Inner and Outer Settings. Using the CFIR model to frame challenges to DMHI implementation yielded useful lessons, especially when peers are engaged as partners in planning and implementation process. Successful implementation will be enhanced by ensuring adequate environmental readiness for tech-based interventions, clear role definition, streamlined peer hiring processes, and well-delineated lines of communication locally and across sites.

Matrix metalloproteinase-14 (MMP-14) and Cathepsin-B (Cat-B) are overexpressed in glioblastoma (GBM) and not normal brain, making them promising targets for prodrug activation. We investigated a novel combination therapy using two tumor-enzyme activatable theranostic nanoprobes (TNP): TNP-MMP-14, which disrupts the blood tumor barrier via MMP-14 activation, and TNP-Cat-B, which selectively targets GBM cells through Cat-B activation. We hypothesized that combining TNP-MMP-14 and TNP-Cat-B would enhance TNP tumor accumulation and therapeutic efficacy compared to TNP-Cat-B monotherapy. Thirty NSG mice with luciferase-expressing GBM39 tumors received either TNP-MMP-14 plus TNP-Cat-B, TNP-Cat-B only, or saline. Magnetic resonance imaging (MRI) was conducted pre- and post-treatment, with T2* relaxation times analyzed using a generalized linear model. Histopathological differences were assessed using Kruskal–Wallis and Mann–Whitney tests. A Bonferroni correction was applied to account for multiple comparisons. Combination therapy significantly reduced tumor T2* relaxation times (12.98 ± 4.20 ms) compared to TNP-Cat-B monotherapy (22.49 ± 3.95 ms, p < 0.001). The apoptotic marker caspase-3 was also significantly higher in the combination group (64.46 ± 23.43 vs. 15.93 ± 5.81, p < 0.001). These findings demonstrate the potential of dual-enzyme activatable nanoparticles to enhance GBM treatment by overcoming drug delivery barriers and improving therapeutic efficacy over monotherapy.

In vivo genetic diversifiers have previously enabled efficient searches of genetic variant fitness landscapes for continuous directed evolution. However, existing genomic diversification modalities for mammalian genomic loci exclusively rely on deaminases to generate transition mutations within target loci, forfeiting access to most missense mutations. Here, we engineer CRISPR-guided error-prone DNA polymerases (EvolvR) to diversify all four nucleotides within genomic loci in mammalian cells. We demonstrate that EvolvR generates both transition and transversion mutations throughout a mutation window of at least 40 bp and implement EvolvR to evolve previously unreported drug-resistant MAP2K1 variants via substitutions not achievable with deaminases. Moreover, we discover that the nickase’s mismatch tolerance limits EvolvR’s mutation window and substitution biases in a gRNA-specific fashion. To compensate for gRNA-to-gRNA variability in mutagenesis, we maximize the number of gRNA target sequences by incorporating a PAM-flexible nickase into EvolvR. Finally, we find a strong correlation between predicted free energy changes underlying R-loop formation and EvolvR’s performance using a given gRNA. The EvolvR system diversifies all four nucleotides to enable the evolution of mammalian cells, while nuclease and gRNA-specific properties underlying nickase fidelity can be engineered to further enhance EvolvR’s mutation rates.

The Gulf Cooperation Council (GCC) countries have experienced rapid coastal development over the past decades, significantly impacting their marine ecosystems. This study aimed to study the land use and land cover change for coastal vegetation in GCC countries from 2000 to 2023 using remote sensing and machine learning techniques, and to identify the impact of climate and anthropogenic factors on coastal vegetation cover of the GCC. We used Landsat satellite imagery and a Random Forest classification algorithm to map various land cover classes along the GCC coastline. Our results revealed significant changes in land cover, as seen by an increase in artificial built-up areas by 15.5% in two decades and a corresponding decrease in bareland and dense vegetation. We also observed an increase in mangrove and seagrass areas, likely due to recent conservation and afforestation efforts. The spatiotemporal analysis showed trends in land cover changes, with agricultural areas generally increasing (21.6%) and bareland steadily decreasing (− 21.2%). Dense vegetation declined by 35% from 2003 to 2023, while mangroves increased by 6%. A multiple linear regression analysis among the various climatic and anthropogenic factors showed that higher temperature positively affected mangroves and seagrass while having a negative relation to dense vegetation. Anthropogenic factors such as urban expansion and agricultural growth negatively impacted dense vegetation. Our findings underscore the need for integrated coastal management strategies balancing economic development with environmental conservation. Further research using higher resolution imagery and advanced classification techniques could improve accuracy and use of the results on a localized level. Our results also provide a baseline for future monitoring and management of coastal ecosystems in the GCC region. Graphical Abstract The significance of this research lies in its ability to quantify the impact of both climatic and anthropogenic factors on vegetation patterns through multiple linear regression analysis. The graphical abstract provides a structured workflow for analyzing vegetation dynamics and land cover changes from 2000 to 2023, integrating remote sensing techniques, machine learning classification, and regression analysis. The left section outlines a methodological framework, beginning with Landsat data collection and preprocessing, followed by the extraction of spectral indices and compositing image collections to ensure consistent analysis. A key step involves training and validating a Random Forest classification model, which is refined through classifier comparison and hyperparameter tuning. The analysis further extends to accuracy assessment, temporal mapping of land cover changes, and climate data integration, enabling a robust evaluation of long-term environmental transformations. The right section presents the spatiotemporal outcomes of the classification, showing mapped land cover classes across the Arabian Peninsula, focusing on urbanization, vegetation shifts, and water bodies. The maps highlight key land cover categories such as mangroves, dense vegetation, agricultural land, and built-up areas, offering critical insights into the effects of anthropogenic activities and climate variability. The integration of multiple linear regression enables a deeper understanding of the relationship between climatic drivers, human interventions, and ecosystem transformations. This comprehensive approach provides valuable tools for environmental monitoring, sustainable land management, and policy planning, particularly in regions vulnerable to climate change and rapid urban expansion.

Carbon dioxide (CO2) removal (carbon dioxide removal (CDR)) that combines decreased greenhouse gas emissions with atmospheric CO2 reduction is needed to limit climate change. Enhanced rock weathering (ERW) of ground silicate minerals is an emerging CDR technology with the potential to decrease atmospheric CO2. However, there are few multi‐year field studies and considerable uncertainty in field‐rates of ERW. We explored combining finely ground metabasaltic rock with other soil CDR technologies (compost and biochar amendments) to stimulate carbon (C) sequestration. The combined ground rock (GR), compost, and biochar amendment had the greatest increases in soil C stocks over 3 years (15.3 ± 4.8 Mg C ha⁻¹). All other treatments slowed or reversed background C losses, with GR‐only treatments reducing rates of soil C loss relative to the control but still losing soil C over time. Ground rock amendments lowered nitrous oxide (N2O) emissions by 11.0 ± 0.6 kg CO2e ha−1 yr⁻¹ and increased methane (CH4) consumption by 9.5 ± 3.5 to 18.4 ± 4.4 kg CO2e ha−1 yr⁻¹; while noteworthy, emissions reductions were an order of magnitude smaller than organic C sequestration with compost amendments. The combined amendment yielded the greatest estimated net ecosystem benefit (3 year relative changes in soil C, estimated ERW rates, and greenhouse gas emissions) of −86.0 ± 24.7 Mg CO2e ha⁻¹. Benefits were dominated by soil organic C gains, directly from organic amendments and indirectly from increased plant growth. Weathering rates were <10% of the theoretical potential. Combined ERW and organic amendments increased estimated weathering rates and stimulated soil organic C sequestration.

Holo‐omics provide a novel opportunity to study the interactions among fungi from different functional guilds in host plants in field conditions. We address the entangled responses of plant pathogenic and endophytic fungi associated with sorghum when droughted through the assembly of the most abundant fungal, endophyte genome from rhizospheric metagenomic sequences followed by a comparison of its metatranscriptome with the host plant metabolome and transcriptome. The rise in relative abundance of endophytic Acremonium persicinum (operational taxonomic unit 5 (OTU5)) in drought co‐occurs with a rise in fungal membrane dynamics and plant metabolites, led by ethanolamine, a key phospholipid membrane component. The negative association between endophytic A. persicinum (OTU5) and plant pathogenic fungi co‐occurs with a rise in expression of the endophyte's biosynthetic gene clusters coding for secondary compounds. Endophytic A. persicinum (OTU5) and plant pathogenic fungi are negatively associated under preflowering drought but not under postflowering drought, likely a consequence of variation in fungal fitness responses to changes in the availability of water and niche space caused by plant maturation over the growing season. Our findings suggest that the dynamic biotic interactions among host, beneficial and harmful microbiota in a changing environment can be disentangled by a blending of field observation, laboratory validation, holo‐omics and ecological modelling.

The purpose of this study is to examine and interpret machine learning models that predict dry eye (DE)-related clinical signs, subjective symptoms, and clinician diagnoses by heavily weighting lifestyle factors in the predictions. Machine learning models were trained to take clinical assessments of the ocular surface, eyelids, and tear film, combined with symptom scores from validated questionnaire instruments for DE and clinician diagnoses of ocular surface diseases, and perform a classification into DE-related outcome categories. Outcomes are presented for which the data-driven algorithm identified subject characteristics, lifestyle, behaviors, or environmental exposures as heavily weighted predictors. Models were assessed by 5-fold cross-validation accuracy and class-wise statistics of the predictors. Age was a heavily weighted factor in predictions of eyelid notching, Line of Marx anterior displacement, and fluorescein tear breakup time (FTBUT), as well as visual analog scale symptom ratings and a clinician diagnosis of blepharitis. Comfortable contact lens wearing time was heavily weighted in predictions of DE symptom ratings. Time spent in near work, alcohol consumption, exercise, and time spent outdoors were heavily weighted predictors for several ocular signs and symptoms. Exposure to airplane cabin environments and driving a car were predictors of DE-related symptoms but not clinical signs. Prediction accuracies for DE-related symptoms ranged from 60.7 to 86.5%, for diagnoses from 73.7 to 80.1%, and for clinical signs from 66.9 to 98.7%. The results emphasize the importance of lifestyle, subject, and environmental characteristics in the etiology of ocular surface disease. Lifestyle factors should be taken into account in clinical research and care to a far greater extent than has been the case to date.

We use supercharacter theory to study moments of Gaussian periods. For $p-1=dk$ and fixed k, we compute the fourth absolute moments for all but finitely many primes p. For d fixed, we relate the fourth absolute moments to the number of rational points on modified Fermat curves. For small d, this relation is in terms of a single curve. For larger d, we provide both exact formulas using families of modified Fermat curves and bounds via Hasse–Weil.

Matrix optimization has various applications in finance, statistics, and engineering, etc. In this paper, we derive the Lagrangian dual of the matrix optimization problem with sparse group lasso regularization, and develop an adaptive gradient/semismooth Newton algorithm for this dual. The algorithm adaptively switches between semismooth Newton and gradient descent iterations, relying on the decrease of the residuals or values of the dual objective function. Specifically, the algorithm starts with the gradient iteration and switches to the semismooth Newton iteration when the residual decreases to a given threshold value. If the trial step size for the semismooth Newton iteration has been shrunk several times or the residual does not decrease sufficiently, the algorithm switches back to the gradient iteration and reduces the threshold value for invoking the semismooth Newton iteration. Under some mild conditions, the global convergence of the proposed algorithm is proved. Moreover, local superlinear convergence is achieved under one of two scenarios: either when the constraint nondegeneracy condition is met, or when both the strict complementarity and the local error bound conditions are simultaneously satisfied. Some numerical results on synthetic and real data sets demonstrate the efficiency and robustness of our proposed algorithm.