22689 research outputs found
Sort by
The Role of Angiogenin within Mitochondria and its Impact on Cell-to-Cell Fusion
No full textCell–cell fusion is essential for a variety of physiological processes and imposes high energy demands on the cell. Although mitochondrial activity is upregulated during fusion, the mechanism coordinating mitochondrial ribosome (mitoribosome) biogenesis to meet these demands has remained unclear. Here, we identify angiogenin (ANG) as a key regulator of mitoribosome biogenesis during myoblast and osteoclast fusion. Upon fusion initiation, ANG translocates from nucleus to mitochondria, where it promotes mitochondrial adaptation and supports the translation of respiratory complex proteins required for ATP production. Loss of ANG impairs mitoribosome assembly, mitochondrial function, and cell–cell fusion, resulting in disrupted skeletal muscle regeneration and defective bone remodeling. Using genetic and molecular approaches, we demonstrate that ANG is essential for mitochondrial remodeling specifically during fusion events. Our findings establish ANG as a regulator of mitochondrial fitness in energy-intensive cellular processes and highlight its physiological importance in muscle and bone homeostasis
Enhancing Survival Prediction Models: Insights on Biomarker Inclusion and Model Updating
No full textPredictive models are essential tools for informing clinical diagnosis and prognosis. When new prognostic biomarkers are identified, researchers must evaluate whether updating existing models justifies the additional cost and complexity of data collection. This study examines how model performance metrics—including calibration slope, C-index, D-statistic, R2 measures, time-dependent Area Under the Curve (AUC), Brier score, and Index of Prediction Accuracy (IPA)—are influenced by changes in sample size, censoring proportion, variable inclusion, and survival distribution assumptions. Using Monte Carlo simulations under exponential and Weibull survival distributions, we assess the impact of incorporating previously omitted covariates and different variable selection strategies on predictive accuracy and calibration. We also provide sample size recommendations to support efficient model updating efforts. Our findings offer practical guidance for developing, updating, and validating predictive models for survival outcomes in clinical research
THE CONVERGING COMMERCIAL LANDSCAPES OF TOBACCO AND CANNABIS: RETAIL, PERCEPTUAL, AND SPATIAL INSIGHTS
Get PDFBackground:
The intersection of tobacco and cannabis retail environments poses public health challenges, particularly as co-use of these substances is linked to increased health risks. This dissertation investigates how the retail availability of cannabinoid products intersects with the tobacco landscape, their combined impact on consumer perceptions through co-packaging exposure, and the spatial distribution of both retailer types.
Methods:
Aim 1 utilized in-person audits of licensed tobacco retailers (n=1,402) across New York City, Philadelphia, and San Francisco in 2023 to assess cannabinoid availability and store-level correlates. Aim 2 employed a randomized online experiment with adult participants (n=602) in 2024 to evaluate the effects of cannabis and tobacco packaging co-exposure on perceptions of tobacco products. Aim 3 analyzed spatial patterns of licensed tobacco retailers (n=6,525) and licensed cannabis dispensaries (n=63) in New York City using point pattern and regression analyses.
Results:
In Aim 1, 9.8% of tobacco retailers sold products labeled as containing tetrahydrocannabinol (THC) or the principal psychoactive constituent of cannabis, with the highest availability in smoke/vape shops (66.7%). Retailers with e-cigarettes or cigars available were significantly more likely to sell cannabinoids, while those with cigarette availability were less likely to do so. Aim 2 revealed that cannabis packaging reduced positive perceptions of cigarettes (e.g., quality, appeal) but increased interest in cigars and lowered perceptions of their harm. Aim 3 found clustering of tobacco retailers around cannabis dispensaries, with commercial zoning strongly correlating with high-density areas for both retailers.
Discussion:
These findings show patterns of convergence between tobacco and cannabis retail landscapes, suggesting this intersection may occur through various avenues including within individual retailers that carry both product types and through spatial proximity of distinct retailers. Findings further indicate that structural factors, such as zoning policies, may inadvertently influence patterns of tobacco and cannabis co-availability. This is worrisome as co-availability patterns can shape consumer perceptions of tobacco, complicating efforts to reduce use. Findings call for integrated regulatory approaches that account for the multiple pathways through which these markets intersect, including shared retail spaces, neighborhood-level co-exposure dynamics, and emerging patterns of product availability that contribute to dual exposure risks
Racial and Ethnic Disparities in Hypertension Treatment Management: Patterns, Consequences, and Interventions
No full textBackground: Racial disparities persist in the United States, with non-Hispanic Black adults experiencing higher hypertension prevalence and lower blood pressure (BP) control rates compared to non-Hispanic White adults. These disparities contribute to increased risks of cardiovascular disease, stroke, and chronic kidney disease. Despite being well recognized, progress in reducing BP control disparities between Black and White adults remains limited.
Objectives: This dissertation examined Black-White disparities in hypertension management by measuring patterns of disparity in treatment intensification, assessing how racial disparities in medication adherence contributed to BP control disparities, and evaluating the effectiveness of a multilevel intervention in reducing adherence disparities.
Methods: First, we examined challenges in measuring disparities and applied the Target Study conceptual model to electronic medical record data from a large Mid-Atlantic health system to measure annual racial disparities in treatment intensification from 2018 to 2022. Second, we analyzed data from the enhanced standard of care plus arm of a pragmatic, cluster-randomized trial, which enrolled participants between August 2017 and October 2019, to estimate the extent to which racial disparities in medication adherence contributed to disparities in BP control. Third, using the same trial data, we evaluated the impact of a multilevel collaborative care team intervention on racial disparities in medication adherence compared to enhanced standard of care plus.
Results: In Aim 1, adjusted analysis showed Black adults had lower treatment intensification rates than White adults, with annual differences ranging from 3% to 4% (e.g., 37% versus 41% in 2022). In Aim 2, the observed BP control disparity was -9.8 percentage points, and if Black participants had the same adherence distribution as White participants, the BP control disparity would be reduced by an estimated 6.0 percentage points. In Aim 3, the multilevel intervention modestly improved medication adherence among White participants but did not improve adherence among Black participants or reduce racial disparities in adherence.
Conclusion: While many questions remain, this dissertation advances methods for measuring Black-White disparities in hypertension management and provides robust evidence on two target factors, treatment intensification and medication adherence, that emerge as potential key strategies for improving BP control and reducing disparities
Cell death resistance in clinical isolates of Cryptococcus neoformans and its relationship to in vitro virulence factors
No full textFungal pathogens are a pressing public health issue that is under-researched and underfunded. With the frequency of infection rising each year and too few antifungal drugs available, the World Health Organization (WHO) called out the need to address and act on the crisis. The knowledge gap for fungal pathogenesis demands us to approach the topic in novel ways. One strategy is to take advantage of intrinsic cell death pathways in fungi to eliminate the infection. Little is known about fungal programmed cell death, however, there is compelling evidence in several model fungal species that provides a basis for considering that distantly related human pathogenic fungi may also encode regulated cell death mechanisms. The aim of this thesis is to investigate if fungal cell death resistance may be a type of virulence factor for human pathogenic fungus Cryptococcus neoformans. Using our novel cell death assays, we tested virulent isolates of C. neoformans from patients for cell death resistance. Results suggest that cell death assays may be a better correlate for virulence compared to the widely used in vitro virulence assays used to measure different aspects of virulence for C. neoformans. These studies also reveal death-resistant and death-sensitive phenotypes within each clinical isolate. This may pose difficulties in identifying genetic determinants of virulence in clinical isolates of C. neoformans by not knowing which subpopulation of cells is most responsible for disease. We propose that cell death assays may serve to identify and focus research on the relevant subpopulations within clinical isolates
Beliefs About Information Across Languages
No full textThe first part of this dissertation concerns information extraction in the cross-lingual setting, wherein data is available in a source language but the system must operate on text in another target language. I examine the downstream impact of a number of methods for projecting training data from the source language to a target language, then propose a model that makes use of the correspondence between the resulting data in the two languages.
The second part concerns beliefs that entities have toward situations, which I view through the lens of the linguistic phenomenon of modality. Text contains a wide variety of beliefs not addressed in standard information extraction, and I argue that it is insufficient to extract just the situations themselves: capturing kinds and strengths of entities' beliefs toward those situations is needed for a more comprehensive understanding of a text. I propose a representation for characterizing beliefs, collect a dataset of entities' beliefs with regard to predicate-argument structures in the Universal Decompositional Semantics corpus, and build models to predict the beliefs, demonstrating the feasibility of annotating data and training models for the task.
In the third part, I combine these two threads by building models that predict entities' beliefs about situations described in text in a language for which human-annotated training data is not available
THE ASSOCIATION BETWEEN SOCIOECONOMIC STATUS AND MACROSOMIA BY FIRST BIRTH IN THE UNITED STATES: PREGNANCY RISK ASSESSMENT MONITORING SYSTEM (PRAMS 2012-2015)
Get PDFIntroduction
Macrosomia is an adverse pregnancy outcome characterized by fetal overgrowth (fetal birth equal to or over 4000 grams or over 4500 grams). Long term complications of macrocosmic infants affect their long term health through increased risk of type 2 diabetes, childhood obesity, hypertension, cardiovascular disease, and low blood sugar. The literature surrounding the main contributors of macrosomia is associated with maternal health conditions and development during pregnancy. However, there is little research that focuses on the socioeconomic factors that would impact the likelihood of macrosomia. This study aims to examine specific socioeconomic factors (education, income, insurance, WIC) and their influence on the risk of macrosomia.
Methods
The Pregnancy Risk Assessment Monitoring System (PRMAS) surveillance data was analyzed, consisting of 107,557 live births that were first born during 2012-2015. Descriptive statistics were used for mother’s characteristics and evaluated using Pearson Chi-squared test between group differences. Logistic regression models were used to examine associations between socioeconomic factors and selected maternal characteristics. All analysis were performed using R (version 4.4.3) and a p-value of <0.05 was considered statistically significant.
Results
Of the 107,557, the overall prevalence of macrosomia was 1.0% (1095 mothers). Macrosomia infant outcomes had a high percentage of women that were in the 30-34 age range (34.3%), who were White (65.8%), with the highest level of education being Bachelors or more (34.0%), and pre-pregnancy BMI of over 30 (42.9%) with gestational weight gain over the recommended guidelines (64.5%). In a multi-model logistic regression, the adjusted models which included age, race, gestational weight gain, and income, did not yield statistically significant results.
Conclusion
Although the study did not reveal a public health significance to address these socioeconomic factors in the context of macrosomia, it is still important to understand how these factors interplay with clinical risk factors of macrosomia which are stronger contenders as a direct correlation to increased risk of developing macrocosmic infants
Essays on the Creator Economy
Get PDFThis dissertation consists of three independent essays in economic theory that broadly touch upon various strategic settings with information frictions in the digital economy.
In Essay I, I investigate how an agent optimally decides when to submit a project proposal when the principal's evaluation is imperfect. Through a tractable model, I identify factors that cause the agent to prematurely submit preliminary work that has not yielded true success (i.e., to \textit{spoof}). I characterize the unique equilibrium of the model, and I investigate the economic implications and welfare consequences of the agent's spoofing. The possibility of spoofing harms both the principal and agent, and welfare losses persist even when the agent can achieve success on the project at an infinite rate or becomes arbitrarily patient. In an extension of the model, I consider a long-term relationship where the agent can resubmit after rejection. The principal can benefit from a long-term relationship if and only if she is sufficiently patient. However, resubmission opportunities make the agent worse off.
Essay II is a joint work with Itay Fainmesser, where we investigate a digital marketing channel: affiliate marketing. In affiliate marketing, retail platforms offer individuals (affiliates) unique \textit{affiliate links} that allow tracking. Our results show that by choosing its affiliate compensation model, a retailer can affect the equilibrium in the market for recommendations in a way that has significant implications for consumer surplus, affiliates' revenues, aggregate welfare, and, ultimately, the retailer's profits. Surprisingly, especially when both consumers and affiliates prefer recommendation conformity, a \textit{conformity equilibrium} may not exist. Consumers' expected search length, affiliates' product information accuracy, and consumers' ability to learn about products' qualities also affect the compensation models that admit conformity equilibria.
In many consulting environments, the expert often assertively recommends the client to take an action but is vague about the probability of that action's outcomes, making the recommendation ambiguous. Essay III (joint with Shubhranshu Singh) analytically investigates this phenomenon by incorporating the client's optimism and attitudes toward ambiguity into a strategic communication model. A primitive premise is that a more ambiguous message can lower the expert's communication cost by freeing the expert from further explanations. In equilibrium, the expert will claim the possibility of a range of probability distributions with a lower end strictly below their precise observation. By choosing an optimal level of ambiguity, the expert trades off the client's expected payoffs for self-benefits from cost reduction and the extra perks from the focal action. When the client cannot exert effort to disambiguate the expert's message, the expert benefits from the client's greater optimism and lower aversion to ambiguous information. Interestingly, when the client exerts costly effort, the expert can leverage the client's ambiguity attitude. As the client becomes more ambiguity averse, precise information about the less preferable option becomes even less preferable---high ambiguity aversion mitigates the ambiguity of the information about the focal option, inducing the expert to send a more ambiguous message
Deep Learning-based Content Creation: Relighting, Synthesis and 3D Editing
Get PDFIn an era where billions of people capture and share digital content, transforming casual captures into aesthetically compelling works often requires extensive manual editing. This dissertation explores deep learning-based approaches to facilitate and automate content creation, enabling high-quality visual content generation within seconds.
The first part of this dissertation focuses on portrait photography, particularly on synthesizing lighting effects—an essential aspect of portrait aesthetics. We develop deep learning techniques for (1) user-friendly interactive lighting editing, (2) creating realistic 3D effects such as shadows and novel views from a single image, and (3) designing a temporally consistent portrait performance relighting system leveraging video diffusion models.
Beyond portrait photography, the second part of this dissertation extends to general 3D scene editing. We explore a novel method for artistically stylizing 3D Gaussian Splatting models, enabling creative modifications to 3D representations.
The final part of this dissertation addresses privacy concerns in deep learning-based content creation. We introduce federated learning-based frameworks that enable high-quality visual synthesis while preserving data privacy. This is crucial for applications involving privacy-sensitive data, such as human faces, where data sharing for model training is prohibited.
Through these contributions, this dissertation aims to bridge the gap between high-quality
content creation and everyday digital capture, making advanced visual editing more accessible, efficient, and privacy-conscious
The Role of KDM4C and CDKN2A on HIF-1 Activity in Ollier Disease and Maffucci Syndrome
No full textOllier Disease and Maffucci Syndrome (OD/MS) are rare disorders called enchondromatoses that are characterized by the presence of benign cartilage tumors in the medulla of metaphyseal bones. Previous work has found candidate causative germline variants in six genes (HIF1A, VHL, IDH1, IDH2, JMJD2C, and CDKN2A) known to be associated with the Hypoxia Inducible Factor-1 (HIF-1) pathway, which promotes cellular adaptation in low O2 (hypoxic) conditions, in around 22% of OD/MS patients. Jumonji Domain-Containing 2C (JMJD2C) encodes for Lysine Demethylase 4C (KDM4C), which demethylates the triple-methylation on lysine 9 of histone 3 (H3K9me3) and acts as both a HIF-1 coactivator and target gene in breast cancer cells. Cyclin-Dependent Kinase Inhibitor 2A (CDKN2A) encodes for tumor suppressor p14ARF, which had previously been found to inhibit HIF-1α activity in U-2OS and Hep3B cells by sequestering HIF-1α in the nucleolus. However, the impact of KDM4C and CDKN2A germline variants on HIF-1 activity in the context of OD/MS remains unknown. Here, we investigated the role of KDM4C, CDKN2A, and their specific variants on the HIF signaling pathway to understand how these mutations may be associated with the development of OD/MS through HIF-1 activity. We performed RT-qPCR, immunoblot assays, and luciferase reporter assays to test HIF-related activity in various cells, including KDM4C knockdown (KD) HEK293T and Hep3B cells, fibroblasts from controls and one patient with KDM4C variants p.Tyr4Cys and p.Ala774Thr, CDKN2A KD Hep3B cells, and HT1080 and JJ012 cells overexpressing CDKN2A WT and variants p.Ala17Gly and p.Ala121Thr. We found that KDM4C KDs showed a significant reduction in HIF-1, but not HIF-2, transcriptional activity in Hep3B cells. We also observed that patient and control fibroblasts showed no significant differences in HIF-1-mediated transcription at hypoxia. Additionally, CDKN2A KD in Hep3B cells did not promote changes in HIF target gene transcription, while CDKN2A overexpression reduced HIF-1 transcriptional activity in HT1080 and JJ012 cells, with no differences between WT and variants p.Ala17Gly and p.Ala121Thr. Taken together, we can conclude that both the KDM4C and CDKN2A germline variants studied did not have significant effects on HIF-1 activity in fibroblasts, HT1080, and JJ012 cells