36214 research outputs found
Sort by
Loss of the USP22 deubiquitylase confers resistance to chemotherapy in small cell lung cancer
No full textThesis (Ph.D.)--University of Washington, 2025Small cell lung cancer (SCLC) responds exceptionally well to cytotoxic chemotherapy. However, relapse with the emergence of chemoresistant disease is rapid and accompanied by poor treatment outcomes. To understand the genetic basis of chemoresistance in SCLC, we applied in vivo CRISPR deletion screening to patient-derived xenograft (PDX) models. Top screen hits included genes encoding components of the transcriptional co-activator SAGA (Spt-Ada-Gcn5 acetyltransferase) complex. We demonstrate that deletion of the SAGA deubiquitylase USP22 conferred cisplatin/etoposide resistance in two chemosensitive PDX models, and that restoring expression in a PDX model harboring homozygous truncating mutation of USP22 re-sensitized tumors to chemotherapy. USP22 loss increased gene body histone H2A-K119 monoubiquitylation in genes encoding key regulators of neuronal differentiation and suppressed neural and neuroendocrine gene expression including targets of ASCL1. Chemoresistance following USP22 loss reflected attenuated DNA damage-driven phosphorylation events and apoptosis, in conjunction with increased expression of glycolysis and hypoxia-related genes. Glycolysis program upregulation may reflect a targetable vulnerability, as inhibition of GLUT1 re-sensitized USP22-null tumors to chemotherapy
Morphologically driven swimming dynamics of the longnose skate
Get PDFSkates propel themselves using winglike pectoral fins that generate undulatory waves along the body. We hypothesize that their skeletal morphology, parallel rows of fin rays composed of mineralized radials, enhances locomotion by passively directing waves through stiffness. Stiffness decreases distally down each fin ray due to bifurcation, and the space between parallel fin rays reduces stiffness along the anterior to posterior axis of the fin. We compared swimming in a live longnose skate (Raja rhina) to the passive motion of a deceased specimen driven by a vertical linear actuator. We used motion tracking to quantify frequency, wavenumber, and amplitude. We investigated whether an undulatory wave can be passively generated without muscle activation, and at what frequencies it most closely resembled live swimming. We observed that for live swimming, small increases in frequency were able to significantly increase swimming velocity, with the skate having a preferred frequency and wavenumber range. This range also matched the observed values calculated from the dead skate’s response, with similar frequencies of maximum amplitude for the live and deceased skate. At these peak frequencies, the output wavenumbers also matched. These results indicated that morphology alone supports wave propagation without active muscle input. Musculoskeletal structure constrains and optimizes the swimming frequency and wavenumber, highlighting the role of passive mechanics in undulatory swimming in the longnose skate
Cellular and Molecular Dissection of the CD8+ T-Cell Response to Kaposi Sarcoma: Implications for Immunotherapeutic Strategies
No full textThesis (Ph.D.)--University of Washington, 2025Kaposi sarcoma-associated herpesvirus (KSHV) is one of two known tumorigenic human herpesviruses and the etiologic agent of Kaposi sarcoma (KS), primary effusion lymphoma (PEL), and multicentric Castleman disease (MCD), malignancies that predominantly arise in the context of T-cell deficiency or dysfunction. There are no approved vaccines, antiviral drugs, or immunotherapies that can prevent or eliminate KSHV infection. The impact of this therapeutic gap is most pronounced in sub-Saharan Africa (SSA), where the burden of KS morbidity and mortality is highest, KSHV is endemic, and HIV-1, the greatest risk factor for KS, is highly prevalent. All other human herpesviruses elicit potent and durable antiviral T-cell responses to control infection. However, studies of circulating T-cell responses to KSHV in the blood suggest that this response is heterogeneous, infrequent, and low-intensity, with the identity of their peptide-MHC targets poorly defined. Given the consistent expression of KSHV in KS tumors, we hypothesized that KSHV-specific T cells would be frequently recruited to KS tumors. We therefore analyzed the TCR repertoire of tumor biopsies from 144 Ugandan adults with KS, 106 of whom were people living with HIV (PLWH). Clusters of T cells with predicted shared specificity for uncharacterized antigens comprised approximately 25% of the T-cell repertoire in the KS tumor microenvironment, representing 4,283 unique αβ TCRs potentially recognizing KSHV- or HIV-encoded peptide antigens. From 25 reconstructed TCRs, we identified two ORF6-specific, HLA-B*45:01-restricted TCRs, one HLA-B*57:03-restricted TCR with specificity for ORF59, and one HLA-A*66:01-restricted TCR recognizing ORF57. The ORF6- and ORF59-specific TCRs were detected in 27-29 and 30 tumors from 14 HLA-B*45:01+ and 10 HLA-B*57:03+ individuals, respectively. Therefore, these TCRs provide the first evidence of a shared, i.e., “public,” T-cell response to KSHV. Primary CD8+ T cells engineered with KSHV-specific TCRs from KS tumor-infiltrating lymphocytes displayed high-avidity, MHC-restricted recognition of two distinct cell types infected with KSHV, B cells (PEL) and endothelial cells (iTIME.219). In parallel, we characterized three novel HIV-specific TCRs within KS tumors from PLWH, two targeting Nef71-79 and one recognizing Vpr34-42, and confirmed the presence of a previously described public HIV-specific TCR recognizing Pol982-990. HIV-specific T cells remained detectable in KS tumors from individuals who had been on antiretroviral therapy for up to five months. Together, these findings demonstrate that KS tumors recruit polyclonal, high-avidity CD8+ T cells recognizing multiple lytic KSHV and in PLWH, HIV antigens. The identification of CD8+ T cells capable of killing diverse KSHV-infected cell types will provide the foundation for rational vaccine design and TCR-based immunotherapeutic strategies to prevent or treat KS
Towards Human-Centered Behavioral Sensing for Student Support
Get PDFThesis (Ph.D.)--University of Washington, 2025Behavioral sensing technologies hold significant potential to enhance human support systems, especially in high-stakes domains such as education and mental well-being. However, existing research often prioritizes model performance, limiting its ability to address real-world needs, adapt to diverse populations, or surface potential risks. Focusing on student support, this dissertation advances a human-centered approach to behavioral sensing by deepening the understanding of students' needs, examining potential harms embedded in current practices, and developing behavioral models that align with Human-Centered Machine Learning (HCML) principles. This work begins with a six-year longitudinal study that combines passive sensing and self-reported data to capture the everyday academic and mental well-being experiences of college students. This is followed by a mixed-method study examining how the transition to online learning during COVID-19—a major life event that disrupted routines, support systems, and access to resources—impacted students with disabilities and mental health concerns—revealing their needs from the onset of the pandemic through the following academic year. These empirical findings motivate a deeper investigation into the ethical risks and fairness concerns surrounding behavioral sensing in real-world deployment. Through both quantitative and qualitative studies, we provide evidence of algorithmic bias embedded in existing behavioral models and uncover broader ethical challenges across the behavioral sensing lifecycle. This work highlights the unique nature of fairness in behavioral sensing, identifies stage-specific vulnerabilities, and surfaces systemic barriers to fair and accountable system design. Drawing on these insights, we propose a reflexive fairness framework and actionable guidelines to support more ethically aligned development and evaluation practices. Informed by these insights, we develop and evaluate three predictive modeling approaches that identify at-risk students as early as the first week of an academic term. These approaches integrate fairness, interpretability, and generalizability as core design objectives. Our findings demonstrate the feasibility of operationalizing HCML in behavioral modeling, while also highlighting key trade-offs and design tensions that emerge in practice. Finally, we discuss open challenges and future directions for building responsible behavioral sensing systems that are not only technically robust but also ethically grounded, inclusive, and attuned to the lived realities of those they aim to support
Beyond Reflex: Nociception, Neural Circuits, and the Transformation of Threat into Memory in Drosophila melanogaster
Get PDFThesis (Ph.D.)--University of Washington, 2025A fly flinches, jumps, runs—not randomly, but in a sequence carved by neurons that have been waiting for danger. For any animal, survival depends on knowing when the world has become dangerous—and reacting fast enough to avoid harm. This ability, called nociception, begins with specialized sensory neurons that detect mechanical, thermal, or chemical threats and ends with circuits that orchestrate escape and shape future behavior. In Drosophila melanogaster, nociception has been dissected in detail in larvae, but the adult system remains largely uncharted: how are its nociceptors wired, how do they drive different forms of escape, and how are their sensory properties tuned to the life of the adult fly? This thesis traces the anatomy and logic of escape in Drosophila melanogaster, revealing how a specific class of abdominal nociceptors initiates a spectrum of behaviors that begin with milliseconds of alarm and end with seconds of altered state.In this thesis, I combine connectomics, optogenetics, calcium imaging, single-cell RNA sequencing, and computational modeling to build a multi-level understanding of adult nociceptor function. I begin with a behavioral screen for neurons capable of driving aversion, focusing on a genetically defined subset of abdominal class IV multidendritic neurons (md). Connectomic reconstruction shows that md neurons project to two distinct downstream pathways: circuits for rapid, reflexive escape behaviors (running, jumping) and ascending circuits for arousal and behavioral modulation. Within the ascending pathway, a pair of inhibitory/peptidergic neurons emerges as a candidate mechanism for regulating both the magnitude and persistence of nociceptive responses, balancing sensitivity with stability.
I next examine the molecular identity of md neurons. Single-cell RNA sequencing reveals a shift from the larval polymodal profile toward a mechanosensory-biased expression program, with strong enrichment of Piezo, ppk, and ppk26, and downregulation of thermosensory channels such as TrpA1 and Painless. Yet functional imaging reveals robust heat responses but no detectable mechanosensory activity, pointing to post-transcriptional regulation or context-dependent tuning.
Together, these findings define the first steps of a complete circuit map for adult Drosophila nociceptors and show how parallel pathways coordinate rapid escape with longer-term state modulation. They also reveal that molecular identity does not always predict sensory function—reminding us that in neural systems, what a cell expresses and what it does are related, but not the same. This work offers both a detailed model for nociceptive control in a compact brain and a broader framework for thinking about how evolution balances urgency, persistence, and sensitivity in the face of danger. We also begin to discuss the ethical implications this type of study might present to us
Racial Bias in Telemedicine: A Within-Subjects Study of Medical Student Attitudes and Affect
Get PDFThesis (Ph.D.)--University of Washington, 2025Health disparities are a matter of grave public health significance. Racial health disparities have complex etiologies and correlates but remain when controlling for other social determinants of health and patient factors such as treatment refusal. Residual disparities reflect differences in provider treatment of White and minority patients. One factor contributing to disparate treatment of minority patients is provider implicit bias—non-conscious biases that alter behavior. However, research suggests that intervening directly on implicit bias may not be effective, indicating that novel directions are needed to understand and address health disparities. Effects of implicit bias on disparities may be clarified by articulating and examining the constructs underlying implicit bias. This dissertation examines intergroup anxiety (anxiety that manifests in interracial interactions in response to negative expectations) as a mediator of the relationship between implicit bias and provider behavior. I first conduct a narrative review to understand the literature related to implicit bias and intergroup behavior. Then, I report on a within-subjects study in which medical student participants (N = 71) interacted with Black and White standardized patients in a telemedicine context. In Aim 1, I conducted preliminary video review in hopes of developing a coding scheme to assess nonverbal behaviors indicative of anxiety. In Aim 2, I used regression analysis to examine the associations between intergroup anxiety and communication behaviors. In Aim 3, I used regression analysis and the Baron and Kenny mediation approach to assess the relationships between implicit bias and intergroup anxiety and the direct and indirect effects of implicit bias on communication behaviors. I report that one of six mediation models tested was significant; observer-rated nervousness was negatively associated with observer-rated warmth. However, as I explore in the results and discussion, low interrater reliability and concerns related to model assumptions indicate that readers should exercise caution in interpreting results. Despite methodological concerns, our preliminary findings indicate weak support for intergroup anxiety as a construct that contributes to disparate provider behaviors. While many health disparities researchers hope that further research will improve health disparities, it is the opinion of this author that most research exploring provider factors that contribute to racial health disparities lacks real-world impact and external validity, and that researchers should focus their work on connecting with communities in deep and intimate ways
Uncovering Hierarchical Cellular Mechanisms: Linking Molecular Regulation and Biological Topology
Get PDFThesis (Ph.D.)--University of Washington, 2025This dissertation investigates biological phenotypes across spatial scales, emphasizinghierarchical and topological features through computational physics and machine learning. At the
molecular scale, coarse-grained simulations revealed diverse conformations of cofilin oligomers
stabilized by disulfide bonds, which regulate actomyosin dynamics via redox-sensitive
modifications. At the mesoscale, mechanochemical simulations and network theory uncovered
topological transitions, or "avalanches," in branched actomyosin networks controlled by Arp2/3,
with machine learning models predicting these events. At the cellular scale, the GRIP-Tomo 2.0
framework integrated synthetic cryo-electron tomography with graph-based learning, enabling
robust protein classification through conserved topological fingerprints under limited data.
Together, these studies demonstrate how topological insights across scales illuminate the
physical principles underlying biological function, highlighting the power of physics-based
computation in complex living systems
Improving Experimental Methods to Capture Real-World Human-AI Perceptions and Interactions
Get PDFThesis (Ph.D.)--University of Washington, 2025AI agents are being increasingly used in production settings, but our understanding of how humansexpect AI to behave, and how AI usage influences human behavior, falls short because of the gap between controlled laboratory studies and real-world usage. In this thesis, I develop methodologies to
shrink this gap and further our understanding of how humans perceive and use AI in practice, and how
we can design more relevant technologies. My methodologies are anchored by the observation that participants with greater task immersion and intrinsic motivation allow modeling more realistic behavior,
and simple manipulation of task settings, domains, and incentives can increase immersion. This thesis
discusses my key contributions to making AI research more relevant to potential downstream users. I first consider the role of AI in collaborative problem solving (CPS) and discover a dearth of openresources for conducting research in human-AI CPS when teams are larger than dyads. I approach this
challenge by developing CPS-TaskForge, a CPS environment generator based on a resource management task, hence resembling real-world problems. CPS-TaskForge enables systematic study of CPS
and open data generation by parameterizing tower defense games, and is thus approachable to laypeople
and intrinsically motivating because the task is fun. Next, I explore how potential risks and harms of AI
assistants are perceived and understood by users by grounding the discussion in procedural document
question answering which has tangible and relatable risks to human evaluators, and recruiting evaluators
who are familiar with the domain of procedural documents. I discover how current human evaluation
techniques fail to account for non-deterministic AI behavior and develop a taxonomy of errors that can
help inform the future development of an AI-powered system. Finally, I examine AI-assisted decision
making behavior and explore the influence of performance pressure, a common environmental factor in production settings that lab studies isolate away from, to further our understanding of the sensitivity
of AI advice taking. My methods illustrate the importance, and potential simplicity, of modeling more
realistic deployment settings while conducting carefully controlled studies
DRAG: Diversity in Retrieval Augmented Generation through the Application of Submodular Functions
Get PDFThesis (Master's)--University of Washington, 2025This thesis applies a submodular approach to the reranking stage of Retrieval Augmented Generation to balance the relevance and diversity of the retrieved documents. After initial retrieval using Contriever, we experiment with submodular functions and a baseline of Maximal Marginal Relevance (MMR), a standard function for balancing relevance and diversity. We apply convex combinations of three approaches: 1) a submodular feature based function using LOG1P concavity and Facility Location, 2) One-Hot Quantization (a quantized modular function with a one-hot feature based function) with manual weights and Facility Location, and 3) One-Hot Quantization with exponential weight decay and Facility Location. We perform hyperparameter selection for the submodular functions and for MMR. We evaluate these on five datasets designed for diversity-focused tasks (news, politics, analogies, etc). We show submodular functions outperform or match MMR's performance in nearly all cases, with recall improvements exceeding 20% (relative difference) in the best case scenario. These results suggest a submodular approach can be effective to improve RAG systems, particularly in diversity-sensitive tasks
DETERMINING THE INTRAMOLECULAR MECHANISMS DRIVING ALTERED CONTRACTION IN THE MYOSIN MUTATIONS E525K AND V606M
Get PDFThesis (Ph.D.)--University of Washington, 2025Force generation in the heart relies on the interaction of myosin and actin filaments, a tightly regulated process where subtle changes can lead to heart disease. Mutations in β-cardiac myosin impact the number of available myosin molecules, their binding to actin, and their ATP utilization rate. Understanding how this family of mutations alter heart contraction requires investigation of myosin at the single molecule, the sub-cellular, and the physiological level. This study investigates the mechanism of altered myosin function in two β-myosin mutations (E525K and V606M). The first project, presented in Chapter 2, uses molecular dynamics simulations of E525K and V606M myosin to highlight a regulatory role for the loop 2 structure in crossbridge binding. The second project, presented in chapter 3, involves a deep investigation of the E525K mutation using stem cell derived cardiomyocyte. Cells and tissues with the E525K mutation showed decreased force generation, consistent with dilated cardiomyopathy; however, single myofibril preparations demonstrated that myofibrils containing E525K myosin can generate more force than wild type under some conditions. These findings underscore the importance of multi-scale studies of myosin mutations. While single-molecule biochemical assays are informative, they may not always reflect the complete picture. As cardiac medicine moves towards personalized treatment, in-depth understanding of how specific myosin mutations alter chemomechanics is vital for designing tailored drugs for cardiomyopathy