22689 research outputs found
Sort by
Beyond Voice: Exploring Minimalist Multimodal Smart Speaker Interaction Powered by Audio-Based Motion Synthesis
Get PDFBeyond conventional auditory interfaces, this thesis investigates the integration of minimalist shape-changing embodiment and audio-based motion synthesis for multimodal smart speaker interactions, embodied by the prototype — Lantern. The research is structured around three questions: the technical architecture for synchronized multimodal responses (RQ1), methods for synthesizing audio features into physical motion suitable for minimalist robots (RQ2), and the impact of multimodal responses on user perception and engagement (RQ3).
Technologically, a systematic multimodal interaction architecture was proposed and successfully implemented, comprising speech processing, large language model-driven dialogue generation, audio-to-motion synthesis, and motion-audio synchronization. The Lantern prototype features a single-degree-of-freedom shape-changing mechanism driven by a servo motor, controlled by a custom circuit board. Evaluation of motion synthesis techniques demonstrated that mapping audio amplitude variations to physical deformation is an effective approach for the minimalist dynamic smart speaker design, providing solutions for RQ1 and RQ2 collectively with the proposed system architecture.
A user study with eighteen participants evaluated Lantern's multimodal interactions in reactive confirmation, proactive nudging, and dynamic conversation scenarios to seek preliminary answers for RQ3. Findings indicate that technical limitations, particularly motor noise and audio-motion mis-synchronization, significantly impacted user comfort and preferences. Conversely, idealized multimodal responses significantly improved engagement, clarity, and animacy perceptions, underscoring that successful multimodal integration hinges on effective synchronization and minimal mechanical disturbance.
Overall, this research constitutes an impactful exploration into minimalist multimodal interaction, demonstrating how carefully synchronized audio-motion responses can substantially enhance embodied voice assistant interactions. Insights from this study provide essential guidance for refining future smart speaker designs and interaction paradigms, paving the way toward more engaging and intuitive embodied voice assistants
Identification of Proteins that Interact with Cellular Retinol Binding Protein 1 (Crbp1) In Cardiac Myocytes and Mouse Hearts
No full textRetinoic acid (RA), the active metabolite of vitamin A, is a critical multi-gene master regulator of cardiac development and function. Despite the established importance of RA signaling in heart physiology, the precise mechanisms governing retinol metabolism in cardiomyocytes remain poorly understood. In this study, we attempt to investigate the interactome of Cellular Retinol Binding Protein 1 (Crbp1), a key intracellular chaperone of retinol, to identify enzymatic partners involved in RA biosynthesis in the heart. Using a combination of Crbp1 overexpression, immunoprecipitation, proximity labeling with TurboID, and mass spectrometry-based proteomic profiling in neonatal rat ventricular myocytes (NRVMs), several candidate interactors of Crbp1 were identified. Notably, Dhrs4, Aldh9a1, and Hsd17b4 were significantly enriched in the Crbp1 affinity pulldown and form a functional network suggestive of roles in retinol and retinaldehyde metabolism. In addition to these candidates, the proteomic analysis revealed sets of previously unrecognized Crbp1-associated proteins including Mlip and Prdm2, broadening the understanding of the cardiac retinoid metabolic landscape. These findings provide new insight into the molecular architecture of retinol metabolism in cardiomyocytes and suggest potential targets for precise therapeutic intervention and development of novel therapies to combat heart failure conditions marked by RA deficiency
Investigating Gonadal Adipose Tissue Remodeling in Ovarian Cancer Progression
No full textOvarian cancer is a highly lethal gynecological cancer with worse patient outcomes in patients with obesity. This cancer is known to metastasize preferentially to adipose-rich sites. However, the role of the gonadal adipose tissue, immediately adjacent to the ovaries, in ovarian cancer progression has not been fully elucidated. We hypothesized that the remodeling of the gAT during ovarian cancer progression facilitated tumor growth, particularly in obesity.
To investigate this question, we injected ID8 ovarian cancer cells intraperitoneally into mice fed on either a standard chow or high-fat diet, and collected their gonadal adipose tissue for analysis. Our results confirmed that diet-induced obesity exacerbated tumor growth. In addition, tumor presence led to a gross anatomical change in the gAT. Immunofluorescence revealed that tumor cells were mainly localized on the periphery of gAT, a niche normally occupied by stromal progenitor cells in tumor-free animals, and macrophages were excluded from the area. Both high-fat diet treatment and tumor development altered the composition of the stromal cells and their metabolome. Notably, cancer led to an increased frequency of matrix fibroblasts in the gAT in vivo and stimulated collagen secretion by stromal cells in vitro. Tumor growth had a similar effect on high-fat diet-fed mice, although it was less pronounced.
These findings demonstrate that ovarian cancer progression significantly remodels the gAT architectural and metabolic environment, and that the high-fat diet modulates the tumor-associated effects. This thesis provides preliminary evidence that understanding stromal-cancer interactions can be important in understanding the metastatic processes of ovarian cancer
Targeting the Cascade-Cas3 interface to investigate Cas3 recruitment in a Type I-B CRISPR-Cas system from Thermotoga maritima
No full textCRISPR-Cas is an adaptive immune system prevalent in most bacteria and archaea. These systems are incredibly diverse in their mechanisms and are divided into many subtypes. Type I CRISPR-Cas systems are the most abundant, accounting for 50% of the systems found in nature and consisting of 9 subtypes. Type I systems employ a multi-subunit CRISPR-associated complex for anti-viral defense (Cascade) to target and bind to DNA. The signature protein of Type I systems, Cas3, an ATP-dependent helicase-nuclease, is then recruited to this complex, which binds single-stranded DNA and degrades it. Most of our understanding of the mechanisms of Type I systems comes from the well-characterized Escherichia coli Type I-E system. However, in recent years, several studies elucidating the mechanisms of other subtypes have been published. In this study, we focus on characterizing the molecular mechanisms of a Type I-B CRISPR-Cas system from Thermotoga maritima. Specifically, we targeted the Cascade-Cas3 binding interface to investigate the process of Cas3 recruitment. Through biochemical and mutational analyses, we identified key factors influencing this crucial step in Type I CRISPR interference. Our findings reveal a complex thermal landscape in this system, with behaviors contrasting to other mesophilic systems – with Cas3 recruitment optimal at 60°C, while both Cascade-DNA binding and Cas3 nuclease activity peaked at 80°C. Metal co-factors like magnesium were seen to be essential for binding, while the presence of nucleotide positively influenced Cas3 nicking without affecting Cas3 recruitment. Targeted mutagenesis at the Cascade-Cas3 interface revealed differential impacts of mutations to the core patch, with mutations on Cas3 causing greater binding impairments. Mutating the HTH-loop showed minimal effects, contrary to observations in other systems. In summary, the findings of this study advance our understanding of Type I CRISPR-Cas systems by revealing the dynamic nature of Cas3 recruitment and activity, opening up avenues for targeted manipulation for biotechnological applications
MISMANAGEMENT OF CELLULAR IRON LEADS TO FERROPTOSIS SUSCEPTIBILITY IN CANCER CELLS SURVIVING CHEMOTHERAPY
Get PDFOnce cancer has metastasized it remains incurable as a result of evolved resistance to nearly all systemic therapies. Our group has documented a non-mutational cancer cell state as an underappreciated mechanism of therapeutic resistance1,2. Cells in this state uncouple DNA replication from cell division resulting in high genomic content and cell size (>40x volume). We observe this resistant cancer cell state induced by multiple different classes of chemotherapy, and in vivo at increased frequencies both in metastatic sites and following chemotherapy. As cancer cells in this state are resistant to conventional therapies that aim to limit cancer proliferation and trigger apoptosis2, we sought to engage cell cycle agnostic forms of cell death as a novel therapeutic strategy for resistant cancer3.
Ferroptosis is a cell cycle agnostic form of cell death that occurs due to dysregulation of cellular iron and unconstrained lipid peroxidation. While ferroptosis is emerging as a vulnerability in resistant cancers, efficacy of ferroptosis induction is highly heterogeneous depending on tissue type, cellular metabolism, and many other factors. We characterized cells treated with chemotherapy and found that surviving cells have high amounts of labile iron regardless of cell line or chemotherapy treatment. We performed single cell RNA sequencing (scRNAseq) of these populations and identified canonical cellular responses to iron toxicity including NRF2 signaling, upregulation of glutathione (GSH) biosynthesis and ferritin expression in all settings. Despite the physiological response to iron accumulation in surviving cells, labile iron was not attenuated, likely due to a relative loss of the PCBP1 chaperone.
By inhibiting GPX4, the only mammalian lipophilic antioxidant enzyme, we show that surviving cells have uniform sensitivity to ferroptosis, regardless of originating cell line or chemotherapy. This labile iron driven reliance on GPX4, and thus vulnerability to ferroptosis, is a unique example of ferroptosis sensitivity arising despite active anti-ferroptotic cellular programs (NRF2 signaling, elevated GSH). These studies will directly contribute to development of ferroptosis engaging therapeutic strategies for resistant cancer cell states, particularly concerning the timing of intervention
BROKEN PROMISES: REDESIGNING STEM AWARENESS & READINESS FOR BLACK & LATINX STUDENTS
No full textEducation has long been recognized as a powerful tool for socioeconomic mobility, guiding individuals out of poverty and toward enhanced health and productivity. However, amidst the acknowledgment of education's transformative power, a troubling disparity persists for Black and Hispanic students, casting doubt on the promise of upward mobility. This discrepancy is particularly evident in the STEM (Science, Technology, Engineering, and Mathematics) disciplines, where Black and Hispanic students are significantly underrepresented compared to their White and Asian counterparts. Despite the pivotal role of college access practitioners in facilitating college degree attainment among ethnic minoritized students, the recurring discourse on the scarcity of Black and Hispanic students pursuing STEM fields underscores a broken promise within the education system.
This paper focuses on the intersection of college access practitioners and the diversification of the STEM workforce, examining the critical role of practitioners in facilitating STEM access for Black and Hispanic students. Employing a qualitative phenomenological approach, this study explores the lived experiences of practitioners who promote STEM engagement among ethnic minoritized students. Through in-depth interviews and thematic analysis, the research examines the unique perspectives, challenges, and strategies employed by practitioners in addressing disparities in STEM access and achievement. The findings shed light on the intricate dynamics of counselor-student interactions, the influence of systemic factors on STEM engagement, and the transformative potential of counselor-led interventions
Use of Artificial Intelligence and Machine Learning in Drug and Biological Product Development: Stakeholder Perspectives on Developing FDA Guidance
Get PDFThe integration of artificial intelligence (AI) and machine learning (ML) into drug and biological product development is rapidly transforming the pharmaceutical landscape, presenting both unprecedented opportunities and significant regulatory challenges. This thesis examines stakeholder responses to the U.S. Food and Drug Administration’s (FDA) Discussion Paper on AI/ML in drug development, with a focus on transparency, bias mitigation, data integrity, validation, and regulatory oversight. Drawing from 69 public comments submitted by pharmaceutical companies, technology firms, regulatory agencies, healthcare organizations, and patient advocacy groups, the study systematically analyzes stakeholder perspectives through thematic analysis.
Key findings highlight widespread support for enhancing transparency and explainability in AI models, balanced against concerns regarding intellectual property protection and innovation incentives. Stakeholders emphasize the need for risk-based, context-specific governance frameworks that distinguish between high- and low-risk AI applications. Data integrity, traceability, and lifecycle management emerge as critical priorities, with growing interest in technologies such as blockchain and federated learning to enhance AI model reliability and patient privacy. The analysis also identifies adaptive regulatory approaches—including dynamic validation strategies, regulatory sandboxes, and real-world performance monitoring—as essential for managing the evolving nature of AI systems.
Ethical considerations, particularly algorithmic bias and inclusivity in clinical trials, are underscored as central to building trust in AI-driven drug development. Stakeholders advocate for fairness audits, bias mitigation techniques, and diverse dataset utilization to ensure equitable AI outcomes. The thesis proposes a framework for AI-assisted clinical trial recruitment to enhance diversity and outlines guiding questions for developers and trialists aimed at promoting ethical and operational best practices.
Ultimately, the findings suggest that while technical advancements in AI are critical, the ethical and regulatory infrastructures supporting their deployment are equally vital. This thesis offers actionable insights for regulators, industry leaders, and policymakers striving to establish cohesive, transparent, and ethically grounded AI governance in pharmaceutical research
STRUCTURAL IMPLICATIONS OF GLUCOSE AND COPPER AVAILABILITY ON GLYCATION PROCESSES IN CHINESE HAMSTER OVARIAN (CHO) CELL-DERIVED PROTEINS
No full textChinese Hamster Ovarian (CHO)-derived cell lines can produce proteins with minimized immunogenic risk for therapeutic use, as they can elicit responses for protein folding and glycosylation patterns compatible with human physiology. These cell lines are used in bioprocessing because they can also produce high yields and are suited for scalable bioreactor production. However, cell culture medium in bioprocessing significantly impacts cell growth, protein productivity, quality of generated proteins, and reducing batch-to-batch variation [1]. Therefore, optimizing the media formulation for cell cultivation is paramount to enhancing the desired product quality of the proteins produced. The studies presented in this project describe the impact of media formulations and cultivation strategies on post-translational modifications (PTMs).
One such PTM is glycation, a non-enzymatic glycosylation process that results in irreversible advanced glycation end-products (AGEs). PTMs can induce conformational changes and alter the net charge of the resulting proteins, making it essential to identify charge variants as a regulatory requirement in bioprocessing. Here, the changes in charge variance from different media formulations and additives were analyzed using a proprietary CHO cell line known to result in high cell growth and AGE formation with undesirable charge variance. This project evaluated the impact of glucose availability and the addition of copper (Cu) (1 µM or less) on product quality in two different media formulations.
The various feeding strategies evaluated showed that glucose availability was a limiting factor for cell growth and AGE production in cultures. Cu addition positively impacted cell growth and productivity under most conditions. Still, the results of this addition were inconclusive for protein charge variance due to glucose overfeeding and an increase in extracellular glutamine, which may be a critical factor that influences deamidation and, consequently, AGE production. Overall, osmolarity was determined to be a crucial factor in influencing cell growth and productivity for the given cell line
Eating Disorder Hospitalizations in the United States: Insights and Implications from the National Inpatient Sample
No full textBackground and Objective
Globally, research shows increasing rates of eating disorders as well as a changing demographic profile of people with eating disorders. Few studies have examined recent nationally representative data on young adult patients with eating disorders in the United States and whether hospitalizations differ by race, ethnicity, sex, and primary payer status. This study identified and characterized differences in eating disorder hospitalizations by race, ethnicity, sex, primary payer, length of stay, and cost per hospital stay. The study also characterized the risk of death for young adult hospitalizations with eating disorder diagnoses by race, ethnicity, sex, and primary payer.
Methods
This study is a retrospective national population-based analysis of the Healthcare Cost and Utilization Project’s (HCUP) National Inpatient Sample (NIS) data from 2016-2019. Hospitalizations with eating disorder diagnoses were extracted from the nationally representative National Inpatient Sample. The study’s primary variable, eating disorders, was characterized overall and by individual eating disorder diagnoses. Additional covariates included sex, race and ethnicity, age, primary payer, total cost of hospitalization, death, and length of stay. Bivariate analyses were conducted. Results of bivariate analyses were used to identify predictors of eating disorders, to characterize total hospitalization charges, and to examine the risk of death. Univariable linear regression of total eating disorder related hospitalization charges identified significant predictors that were included in a multivariable regression model to assess associations between hospitalization charges and covariates further and to adjust for potential confounders.
Results
Characteristics of hospitalizations differed by each eating disorder diagnosis code, suggesting differences in diagnosis and treatment of eating disorders by subtype and demographic characteristics. All eating disorder (MDB-010 code) related hospitalization charges showed an association between higher total hospital charges and being of male sex, Hispanic ethnicity, Asian race, Black race, and longer length of stay. This study’s results also showed a significant association between male sex and the risk of death during hospitalization with an eating disorder diagnosis.
Conclusion
This research study identified and characterized hospitalizations for patients with eating disorders ages 18-30 years of age, overall and by specific eating disorder diagnosis, examined the costs per hospitalization and examined the risk of death for eating disorder hospitalizations. Overall, this study provided evidence of diversity in eating disorder presentation in the United States and possible disparities in hospitalizations and hospitalization outcomes for different groups of people with eating disorders. These results can inform clinicians and public health experts about health disparities in eating disorder care in the US, as well as provide support for increased public health surveillance of eating disorders in the United States. Results may help guide additional research efforts into understanding the mechanisms that may place some populations at disproportionate risk, which subsequently could lead to refined and tailored eating disorder prevention and treatment approaches
EPIGENETIC MECHANISMS DRIVING COLLECTIVE EPITHELIAL INVASION IN LUMINAL BREAST CANCER
No full textThe iterative nature of mammary branching morphogenesis suggests an inducible program of differentiation that is most readily accomplished by epigenetic regulation, through the intricate interplay of chromatin-binding proteins. These same developmental programs can be hijacked to promote a malignant state, thereby causing a group of adherent epithelial cells to gain motility and collectively invade from a solid tumor. However, the epigenetic mechanisms underpinning these processes remain incompletely understood. Previous studies have determined that cells at the leading edge of breast cancer invasion exhibit a distinct gene expression pattern compared to those in the bulk of the tumor. These leader cells express basal epithelial genes, including cytokeratin-14 (Krt14) and tumor protein 63 (Trp63). To understand how cancer cells switch between luminal and basal states, we tested the requirement for histone acetyltransferases in both mammary branching morphogenesis and models of luminal B breast cancer. Specifically, through RNA-seq and FACS analysis, we identified that p300/CBP regulate the cell state transition from luminal to basal cell identity by promoting the expression of basal signature genes. Mechanistically, we identified that p300/CBP-mediated acetylation of H3K27 recruits key transcription factors, including Trp63, to promote the invasive program. Our results provide a mechanistic understanding of the epigenetic underpinnings of the basal cell state transition required for luminal breast cancer invasion and provide a pre-clinical rationale for a novel therapeutic strategy