22689 research outputs found
Sort by
CELL DEATH REGULATION AND STRESS-INDUCED CELL DEATH RESISTANCE IN YEASTS
No full textFungal pathogens encounter various environmental stresses during infection, including high temperatures and oxidative bursts produced by host immunity. While there is increasing research on stress resistance mechanisms, how these responses vary in natural and clinical populations remains unclear. This thesis investigated stress-induced cell death in Saccharomyces cerevisiae and Cryptococcus neoformans and explored whether these traits might provide insights about virulence.
As shown previously in S. cerevisiae, deletion of any one of the four AP-3 complex subunit genes leads to significantly enhanced survival under heat-ramp-induced cell death conditions. One or more of the proteins that are transported by AP-3 to the vacuole membrane may be a cell death effector of the AP-3 death pathway following stress. SLM4 is among these candidates. Here, I show that reintroduction of SLM4 into the knockout strain restored cell death sensitivity, further suggesting a role for Slm4 in regulating stress-induced cell death. Further investigation identified CNAG_02068, a homolog of an AP-3 cargo protein in C. neoformans, whose deletion similarly resulted in mild resistance to heat-ramp stress, supporting a role for AP-3 trafficking pathway-mediated regulation of cell death pathways in C. neoformans. Clinical isolates of C. neoformans, assumed to be virulent, were also screened for their responses to heat-ramp and oxidative stress. Among the eight clinical isolates tested, six showed evidence of stress-induced cell death resistance. In these six clinical isolates, either the entire population showed resistance or heterogeneous phenotypes were observed, from which resistant subpopulations could be isolated after further isolation.
In vitro markers of virulence factors were also evaluated in all clinical isolates, including capsule size, melanin production, urease activity, biofilm formation, and phospholipase activity. Although these traits varied substantially among samples, stress-induced cell death resistance was a relatively consistent and widespread feature. Notably, resistance to heat-ramp and oxidative stress did not correlate consistently with changes in virulence factors, suggesting that stress-induced cell death resistance may represent a more conserved and prevalent trait among clinical isolates compared to classical virulence factors
Rules for Thee, But Not For Me: Qualitative Case Studies in Online Health Communication and Content Moderation
No full textHealth misinformation poses a major challenge to effective public health communication. The World Health Organization declared an “infodemic” of false and misleading information complicating health communication during the COVID-19 pandemic. While the pandemic has now ended, the conditions that fueled the infodemic remain. The modern media ecosystem is built around social media platforms, which present both new opportunities and challenges for health communication. This dissertation asked, “What are the implications of positioning social media as both the source of and solution to the infodemic?”
This included three separate qualitative analyses. First, a content analysis of content moderation policies from ten social media platforms. Platforms applied differential consequences based on subjective levels of “harm”: content associated with more severe harms was removed from platforms; content associated with less severe harms was addressed through algorithmic interventions that left content online but reduced its visibility. Next, a secondary qualitative analysis of interviews (n=38) with health communicators working during the COVID-19 infodemic. Participants voiced ambivalence about social media as a site for health communication. Increased transparency from platforms may address some challenges but cannot reconcile tensions between commercial platform values and the mission of health communication to serve the public. Finally, a critical discourse analysis of dual governance models for social media. Users were encouraged to become “creators” to reap the benefits of these programs, but the unequal distribution of risks allows platforms to gain more “advertiser-friendly” content while shifting the burden of content moderation onto creators. “Sensitive content” categorizations are vague, and creators may self-censor health content. These policies serve to uphold the economic and regulatory status quo.
Altogether, these studies advance an argument for an ecological understanding of the infodemic. Applying the lens of critical media ecology opens new avenues for critique but also the possibility for change. Resist the “logic of opacity” that attempts to pass off technological systems as a natural solution to health misinformation. These tools have reinforced existing power asymmetries between platforms and the public, boosting user engagement (and platform revenue) while insulating platforms from external oversight. We conclude by advocating “Rules for All.
Comparing Optical and Interferometric Synthetic Aperture Radar-Based Remote Sensing Methodologies for Conflict Damage Assessment in Lebanon: A Case Study
Full text linkVOICES FROM THE STARS, STORIES FROM THE EARTH
Full text linkThis thesis explores the intersection of science communication, personal narrative, and historical analysis to examine how public understanding of science is shaped through stories. Using a blend of essays, profiles, and interviews, the work navigates themes of space exploration, environmental conservation, and equity in science. Central to the thesis is a focus on how individuals—scientists, communicators, and marginalized voices—advance public engagement with science and shape its narratives. Ultimately, this work underscores the importance of storytelling in making science accessible, relatable, and transformative. By weaving historical, personal, and scientific narratives, the thesis advocates for a more inclusive and informed dialogue about science and its role in shaping our shared future
Towards Refined Proteomics: Spatial and Comprehensive Characterization of Proteins
No full textProteomics was first coined by Marc Wilkins in 1995 in analogy to genomics. Unlike the genome, which remains relatively constant, the proteome is highly dynamic, constantly adapting through biochemical interactions with the genome and environmental perturbations. This dynamic nature gives rise to a diversity far exceeding the 20,000 to 25,000 coding genes, enabling proteins to execute extensive and interconnected cellular functions. Modern proteomics is the comprehensive evaluation of proteins to gain insights into biological processes, whether in a state of health or disease. Many aspects are investigated in this large-scale study of proteins, particularly their structures and functions, including detection, identification, measurement of their concentration, investigation of their spatial/temporal distribution, characterization of modification, characterization of protein-protein interaction and regulation. These interconnected layers of information form a complex and dynamic picture of the proteome. Advances in mass spectrometry-based technologies allow us to perceive this masterpiece with increasing clarity, revealing broader perspectives and finer details. As these technologies continue to evolve, the intricate nuances of this complex landscape will be brought into sharper focus, uncovering previously unseen dimensions of the proteome.
Building on these advancements, this thesis presents significant contributions to the field of proteomics, focusing on the development of workflows, platforms, and methodologies to address critical challenges in protein modification profiling and spatial proteomics. The first major contribution of this thesis is the development of an integrated workflow for the comprehensive characterization of multiple post-translational modifications (PTMs), including phosphorylation, glycosylation, acetylation, and ubiquitination. The second contribution is the development of a novel technology, termed SPOT, for on-site tissue protein labeling tailored to spatial proteomics. Finally, this thesis advances SPOT by transforming it into an automated, precisely controlled printer, paving the way for improved efficiency and accuracy in spatial proteomic studies. The future direction of this thesis addresses the utilization of multiple proteases—trypsin, AspN, and GluC—to improve protein coverage. This approach, combined with the application of TMT-18 labeling, expands the current plexing capabilities from 18 to 54, enabling higher throughput and deeper proteomic analyses
ACTIVE SENSING AND TASK CONTROL IN TWO DISTINCT ANIMAL SENSORIMOTOR SYSTEMS
Full text linkMammals excel at performing complex motor actions, a capability that modern
robotic systems struggle to replicate. This ability arises from a combination of
feedback, feedforward, and active sensing control mechanisms. Feedback control uses external inputs and the current motor state to make adjustments toward a goal state. Feedforward control, on the other hand, maps a desired goal state to motor actions required to achieve the goal state. Active sensing allows animals to gather more information about their environment, often at the cost of goal-directed motor
actions, creating a trade-off between the two. This dissertation explores two distinct sensorimotor systems — visuomotor tracking in humans and echolocation regulation in bats — to understand how these three control mechanisms are implemented.
The first study examines people with cerebellar damage, which impairs motor
control. Using system identification techniques and a virtual reality tracking task, we investigate how feedforward and feedback pathways contribute to motor control and the cerebellum’s role in incorporating external stimuli into motion planning. Our findings show that cerebellar damage induces delays in both control pathways, but it does not alter the overall control dynamics. However, cerebellar damage significantly
impairs the ability to incorporate models of predictable stimuli to improve motion
control.
The second and third study investigate the active sensing behavior of bats
echolocating while hunting prey and characterize active sensing as a closed-loop control
process. They also examine the impact changes in goal states and environmental
factors have on echolocation behavior. In the second study, bats track a moving target in cluttered and uncluttered conditions. In the third study, we incorporate system identification techniques to quantitatively model the relationship between echolocation
changes and target motion by having the bats track oscillatory targets. Our results
demonstrate that bats dynamically adjust their echolocation calls in response to target movement, with environmental clutter inducing global shifts in call parameters. Bats also preferentially tune their echolocation to emphasize higher motion frequencies and to predict target motion.
The combination of these three experimental paradigms provides insights into the biological mechanisms underlying goal-directed motor control and the modulation and implementation of active sensing during motor control
Viscoelastic Behavior and Energy Absorption of Main-Chain Liquid Crystal Elastomers-Based Architected Materials
No full textEnergy absorption plays a crucial role in designing materials for impact mitigation and vibration isolation. Architected materials that harness geometric instabilities, such as snap-through buckling, offer promising capabilities for reversible energy trapping. However, their energy absorption often remains unaffected by strain rates. Incorporating rate-dependent dissipation mechanisms like viscoelasticity can enhance performance, but this approach remains largely unexplored. Liquid crystal elastomers (LCEs) present a unique opportunity in this context. These materials consist of crosslinked polymer networks embedded with stiff, rod-like mesogens, creating a coupling between network deformation and mesogen reorientation. This interaction enables reversible actuation, rate-dependent soft stress behavior, and improved energy dissipation. Compared to conventional elastomers, LCEs exhibit elevated loss factors across various frequencies and temperatures and significant hysteresis under cyclic loading. Both viscoelasticity in the polymer network and soft stress from mesogen rotation contribute to energy dissipation.
This dissertation examines how the combined effects of viscoelasticity, soft stress behavior in LCEs, and energy-trapping mechanisms in instability-based architected materials enhance energy absorption.
The first part investigates the combined effects of viscoelastic dissipation in LCEs and snap-through buckling in a lattice structure with tilted beams under compression. Experiments and finite element simulations using a large-deformation viscoelastic model demonstrated that energy absorption density increased with strain rate. Stacked LCE structures exhibited nonuniform buckling and localized load-unload cycles, amplifying energy dissipation as the number of layers increased.
The second part explores the finite-strain viscoelastic behavior of monodomain LCEs under uniaxial tension when initial mesogens were perpendicular to the load. Simulations and experiments revealed that the more crosslinked LCEs contributed more to elastic strain energy and polymer network dissipation, while mesogen rotation remained the dominant energy absorption mechanism in both crosslinking densities.
The final part examines soft stress behavior from mesogen rotation in LCEs as an additional energy dissipation mechanism in tensile members of structures with tilted beams under compression. Simulations and experiments showed that incorporating horizontal LCE stretching increased energy dissipation by a factor of 2-3 compared to structures with rigid horizontal bars.
This dissertation provides insights into applying LCEs in energy-absorbing systems, paving the way for next-generation impact mitigation structures
A Framework for Respiratory Sound Analysis in Clinical Settings
Full text linkRespiratory diseases remain a major public health concern worldwide, accounting for a significant portion of morbidity and mortality. Pediatric populations are especially vulnerable, with conditions such as pneumonia considered the leading cause for children under the age of five worldwide. While often preventable, early detection and accurate classification of respiratory infections are essential to enable timely intervention and improve patient outcomes. Computerized auscultation analysis (CAA) offers a promising avenue for efficient and precise respiratory disease diagnosis, yet it has not seen widespread clinical adoption due to several persistent challenges. Variability in recording conditions, inconsistencies across diverse healthcare settings, class imbalances in auscultation datasets, and automated models that struggle to capture the uncertainty observed in expert assessments all limit the performance and reliability of current methods.
In this dissertation, we introduce a suite of approaches aimed at overcoming these barriers and advancing CAA toward clinically impactful applications. The core focus lies in detecting wheezes and crackles in pediatric pneumonia. A foundational element is a novel auscultation quality metric that integrates psychoacoustic features with data-driven embeddings, ensuring consistent and reliable signal assessment even under challenging noise conditions. We explore the interplay between model confidence and physician expertise, shedding light on how automated systems can approximate the nuanced uncertainty in clinical diagnoses. We further establish the clinical relevance of these methods through cross-modal validation with chest radiography, demonstrating strong alignment between automated auscultation outputs and conventional diagnostic techniques. Addressing the issue of class imbalance for rarer pathological sounds, we propose constrained synthetic minority oversampling tailored to crackle generation validated by rigorous statistical analysis. Additionally, we introduce robust feature extraction and implement domain adaptation strategies to ensure consistent performance in varying recording environments.
These building blocks are integrated into an end-to-end framework that incorporates confidence-based classification, advanced augmentation techniques, and robust feature extraction. By providing reliable detection of adventitious respiratory sounds in real-world clinical scenarios, this comprehensive system paves way for more accessible and scalable CAA deployments. Through the methods presented in this dissertation, we aim to contribute meaningfully to global efforts in mitigating the burden of respiratory diseases and ultimately improving patient care
Targeting Ras Neoantigens with Novel Cancer Immunotherapies
No full textDeveloping targeted therapies for cancer remains a major challenge in cancer research. Immunotherapies have emerged as an effective tool to target tumor-associated antigens, usually cell surface molecules that are overexpressed compared to normal tissues. However, new strategies for targeting tumor-specific antigens exist. Mutation-associated neoantigens (MANAs) are peptides derived from intracellular mutant proteins that are presented on the cell surface in peptide-HLA complexes. Targeting MANAs offers a strategy to eliminate cancer cells harboring driver mutations encoding intracellular proteins that are otherwise difficult to target with small molecules or immunotherapies. Here, we present work developing a novel TCR mimic bispecific antibody that redirects the immune system against cancer cells presenting a highly recurrent KRAS MANA. We show that this bispecific antibody can specifically activate T cells to eliminate cancer cells harboring this KRAS mutation. We further characterized binding of this bispecific antibody to the KRAS MANA using cryo-EM to elucidate mechanisms of binding specificity and understand how an antibody can effectively differentiate between single amino acid differences in presented peptides. Taken together, this work demonstrates that MANA-targeting bispecific antibodies (MANAbodies) are a promising strategy to target highly recurrent mutations in intracellular proteins, allowing for more specific cancer therapies in the future
One VAE to Squeeze Them All: Causality-Aware Spatiotemporal Compression for Multi-Contrast MRI Latent Unification
No full textMedical image synthesis supports clinical decision-making by enabling tasks such as multi-modal contrast prediction, longitudinal tracking, and lesion characterization. In clinical practice, acquiring all MRI contrasts for each patient is often impractical due to time, cost, and patient burden, making the prediction of missing contrasts a valuable tool. Meanwhile, the heterogeneous nature of MRI data—with both static 3D and dynamic 4D images poses additional challenges for unified modeling. Recent advances in generative models, particularly Variational Autoencoders (VAEs), offer a promising direction for learning compact and informative latent representations from such complex data. In this work, we introduce a novel 3D/4D VAE-GAN framework for multi-contrast MRI image prediction and reconstruction. Our hybrid model preserves high-quality reconstruction and establishes a smooth latent space while dynamically adapting to varying temporal dimensions in MRI sequences. Both quantitative and qualitative evaluations confirm excellent reconstruction quality and clear separation of different lesion subtypes, demonstrating the model’s ability to capture clinically relevant spatiotemporal features. Furthermore, the compressed latent representations can be integrated into diffusion models for image synthesis and missing contrast prediction, and facilitate downstream tasks like lesion segmentation, patient retrieval, and enabling privacy-preserving data sharing. These findings also offers a robust foundation for future data-driven healthcare applications