36214 research outputs found
Sort by
Synthesis, sequencing, and surveillance applications of xenonucleic acids
No full textThesis (Ph.D.)--University of Washington, 2025The four nucleotides that form DNA (A, T, G, C) are central to the storage and propagation of genetic information. While nature has limited itself to four letters, these are not the only nucleotides capable of supporting the genetic code. Unnatural base pairing xenonucleic acids (ubp XNAs) are chemically synthesized nucleic acid analogs that base pair orthogonally to those found in nature. Ubp XNAs expand the accessible biochemical space of DNA, enabling innovation in diagnostics, therapeutics, and catalysts. Despite these advancements, the tools available for ubp XNA synthesis and sequencing remain decades behind those available for standard DNA. In this thesis, we develop tools for the synthesis and sequencing of ubp XNAs. We successfully demonstrate an enzymatic method for inserting a singular ubp XNA into a standard DNA context, and use this method to enable the sequencing of DNA with up to 12 letters. Next, we show that even without these tools, ubp XNAs can be used to overcome existing barriers in standard DNA biotechnologies. We develop an assay that uses two sets of non-standard bases to decrease off-target reactions in a multiplexed polymerase chain reaction (PCR). We benchmark the ability of this assay to detect 20 targets in three different matrices. Finally, we develop a proof-of-concept nucleic acid-based lateral flow platform for pathogen detection in low-resource settings that will integrate ubp XNAs to prevent non-specific hybridization. Together, these studies advance tools for ubp XNAs and demonstrate their abilities to support a new generation of biotechnologies
Signaling properties of asymmetric spider webs
No full textThesis (Ph.D.)--University of Washington, 2025Spiders rely on vibrations transmitted through their webs to mlocate and classify prey, and spider will adjust their behavior according to the gained information. Yet, the cues spiders use to interpret these signals, especially from the initial impact, remain poorly understood. These webs are lightweight, mechanically nonlinear structures that serve as extended sensory systems. Understanding how the design of such systems affects signaling can offer insight into sensing in engineering sensing systems, in the field of for example structural health monitoring, feedback control systems, and optimal sensor placement. This dissertation investigates how specific vibrational cues—those that spiders can realistically measure—contribute to prey localization and classification. A central question is whether structural irregularities, such as eccentricity and narrowness, enhance these cues. The study also examines whether such cues remain reliable despite changes in web design and external parameters like prey mass and impact location, which is critical for robust sensing in systems operating with limited information, like spider webs. Previous research on spider web dynamics has primarily focused on idealized circular webs and local geometric features, overlooking how global web shape and prey mass influence vibrational behavior. While several promising mechanisms for prey detection and localization have been proposed, it remains unknown whether these cues remain reliable across the diverse range of web shapes seen in nature and under varying prey sizes. Critically, spiders must act based on limited local information, measured only at their leg positions, so any effective cue must be robust to changes in design and external conditions without relying on global knowledge of the web dynamics or prey. The study combines a numerical and experimental investigations to explore how spider webs design affects vibrational cues. Spiderweb-like structures were fabricated with biologically accurate tension gradients and tested under dynamic loading to analyze their vibrational behavior, enables with high speed cameras. These experiments were complemented by numerical simulations using the Finite Element method. This dissertation presents three main discoveries. First, a robust pitching mode was identified that allows spiders to localize prey using directional vibrations, independent of web irregularities. Second, a specific cue was found to reflect prey mass through system dynamics. However, ambiguity arises due to overlap between the effects of prey mass and distance to impact, and introducing geometric irregularities actually increases this ambiguity. Still, these cues support robust classification of prey type based on mass. Third, building on these biological insights, the thesis introduces a design method for 3D-printed networks with programmable tension gradients and develops Directional Digital Image Correlation (D-DIC), a displacement measurement technique for optical methods that enables full-field experimental modal analysis from a single impact. This includes tracking edges, not just at intersections, as was possible with DIC. Engineering sensing systems are typically designed for structures with known geometry and dynamics. Sensors can be placed with flexibility, and external perturbations are often treated as non-intrusive to the system's behavior. In contrast, spiders must sense and interpret vibrations in webs with unknown and variable properties. They are limited to sensing vibrations near the web's center, and must rely on cues that remain informative despite variation in design, prey mass, and impact location. Moreover, in spider webs, an impact event significantly alters the system's dynamics, making sensing inherently more challenging. This work shows that spiders overcome these limitations by relying on robust structural cues. These findings offer a foundation for designing sensing systems that perform reliably in uncertain and constrained environments, where full modeling is not feasible and disturbances cannot be ignored
Interpretable Machine Learning for Biomarker Identification in RNA Seq Cancer Data
Get PDFThesis (Master's)--University of Washington, 2025Existing research on RNA Seq gene expression biomarkers has provided various methods to select a small list of genes as cancer biomarkers from a large number of gene expression data. Previous methods for identifying potential gene expression cancer biomarkers have focused on statistical analysis, but other methods have incorporated machine learning, often including Interpretable Machine Learning (iML) techniques. On 16 cancer types from TCGA data, we used inherently interpretable machine learning models: Logistic Regression, Random Forest, and Linear Support Vector Machine to narrow down subsets of potential genes as biomarkers using the trained models' feature importance rankings. We subsequently applied model-agnostic iML techniques, such as Shapley Additive Explanations (SHAP) and Permutation Importance, to narrow down the subsets even further. We compared classification performance between machine learning models trained on iML selected features with features selected by statistical methods, and biomarkers from external research. We found that iML biomarker selection methods lead to comparable or better classification performance on these datasets than the biomarkers from outside research, or from statistical analysis alone. Mutual Information estimation (MI) was a surprisingly useful technique for initial feature selection, and iML techniques improved the MI selected features for classification. We cross-checked potential biomarkers with biomedical annotations and gene pathway analysis, finding some support for the validity of the biomarkers
Little Buoys, Big Ocean: Observations of Wave-Driven Transport of Buoyant Objects in the Nearshore and Marginal Ice Zone
Get PDFThesis (Ph.D.)--University of Washington, 2025A variety of objects drift on the ocean surface, including plastics, sea ice, search and rescue targets, oil, and marine organisms. Accurately predicting their trajectories is essentialfor both scientific and operational applications. The processes that drive the transport of
these objects, particularly on shorter time and space scales, are not fully understood. This
work addresses this problem, particularly investigating the wave-driven transport mechanisms in the nearshore and marginal ice zone. These regions are dynamically complex, and
wave transformation is a dominant feature of both regions. This work primarily uses in situ
observations to investigate buoyant object transport. The observations are from two large
field experiments: the US Coastal Research Program funded During Nearshore Events Experiment (DUNEX) and NASA's Salinity and Stratification at the Sea Ice Edge (SASSIE)
campaign. The following work consists of the development and nearshore deployments of a
small-scale free-drifting wave buoys called microSWIFTS, observations of surfing transport
from the microSWIFTs, and observations of transport driven by wind and waves in the
Arctic marginal ice zone. The microSWIFT is a small buoy equipped with a GPS module to measure the buoy'sposition and horizontal velocities and an Inertial Measurement Unit (IMU) to directly measure the buoy's rotation rates, accelerations, and magnetic heading. Measurements were
collected over a 27-day field experiment (DUNEX) in October 2021 at the US Army Corps of Engineers Field Research Facility in Duck, NC. The microSWIFTs were deployed as a
series of coherent arrays, meaning they all sampled simultaneously with a common time
reference, leading to a rich spatial and temporal dataset during each deployment. Measure-
ments spanned offshore significant wave heights ranging from 0.5 to 3 meters and peak wave
periods ranging from 5 to 15 seconds over the entire experiment. Observations of surfing transport are made using the data collected from the microSWIFTsas part of DUNEX. Surfing events are observed in the drift trajectories of the buoys as 'jumps'
in the time series of cross-shore position. There are 3,172 surfing events observed, with a
median jump amplitude of 8.3 meters and a median duration of 2.5 seconds. The buoy's
trajectories are simulated using three models of increasing physics complexity: 'Wind-Only,"
'Wind and Waves," and 'Wind, Waves, and Surfing." The surfing process is represented using
a probabilistic parameterization. The accuracy of the simulations is significantly improved
when surfing is included. Sea ice in the marginal ice zone is generally considered to be in "free-drift" and acts like adrifter following the wind, waves, and currents. The marginal ice zone is the dynamic region
of sea ice between 15% - 80% sea ice area coverage. A dominant feature of the marginal
ice zone is the attenuation of waves propagating from the open ocean, making this region
analogous to the surf zone. SWIFT drifters were deployed in the marginal ice zone, and their
drift speeds were measured. A combination of wave radiation stress gradient-driven currents,
direct windage on the drifters, and Ekman transport accounts for their observed drift speeds.
These results suggest that wave-driven transport mechanisms cause a significant portion of
the observed drift at the outer edges of the ice pack. Wave-driven transport in the marginal
ice zone may play an important role in shaping the future evolution of the sea ice edge, as
earlier melting and later refreezing create more open ocean and allow larger waves to interact
with the ice edge
Advancing Digital Health Equity in a Safety-Net Health System: Identifying Barriers, Evaluating Training, and Assessing Impact on Diabetes Outcomes
Get PDFThesis (Ph.D.)--University of Washington, 2025Digital health technology, including patient portal use and telehealth visits, has been increasingly utilized across healthcare settings, transforming how individuals access healthcare and contributing to improved quality of care. However, there are differences in which types of patients use digital health technology, stemming from multi-level factors at the structural, contextual, and individual levels. Without better evidence, improved methodology, and proactive interventions to reduce inequity and promote equity, these disparities will persist as the digital divide widens. This dissertation is centered on the experience of individuals at San Francisco Health Network, an urban safety-net health system. Across three papers, I employ novel qualitative and quantitative methods to address critical questions about marginalized populations and digital health technology use, thereby filling gaps in the literature and advancing digital health equity. In the first paper, I employ mixed methods to examine fundamental skill and usability barriers to digital health technology use. Notably, I highlight critical gaps in digital literacy, particularly in device navigation and processing complex tasks, that prevent effective use of these tools, and the need for usability-driven improvements to reduce digital barriers. The second paper uses a zero-inflated negative-binomial generalized linear mixed model to evaluate the impact of person-centered digital training on patient portal uptake and use, considering sociodemographic factors, clinical characteristics, and digital engagement. I identify that patients who participated in tailored training saw a 91% relative increase in average monthly portal users, compared to a 12% relative increase among those who received basic digital support only. More specifically, the basic digital support program especially benefited Spanish-speakers who demonstrated an 80% increase in login counts, although the tailored training significantly benefited participants with low baseline engagement and resulted in a fivefold increase in login frequency compared to pre-intervention rates. The final paper leverages the widespread adoption of digital health technology and remote care engagement in health systems and uses a linear mixed-effects model to examine how combined in-person and remote care utilization patterns impact longitudinal changes in A1c control, and whether these patterns differ across key sociodemographic factors. I found that multiple remote and in-person care utilization patterns were associated with modest but clinically meaningful differences in glycemic control. Specifically, the degree of A1c improvement followed a clear gradient across care patterns, with the least improvement among patients with little or no care and progressively greater gains as remote and in-person modalities were combined, underscoring the value of hybrid engagement for chronic disease management. However, these associations varied by sociodemographic characteristics, revealing disparities in access, adoption, and effectiveness of in-person and remote engagement across patient groups.
Together, these papers outline the growing issue of disparities in digital health technology use and pinpoint solutions and evidence to support accessibility and equitable use. These papers highlight the skills needed to use digital health technology effectively, the role of digital skills training in promoting the use of these tools, and the benefits of engaging in care remotely on patient health outcomes. However, a common thread across all these papers is the persistent presence of disparities and barriers among marginalized populations, and how the influence of these barriers extends beyond disengagement with digital health technology but also affects health outcomes. In my concluding chapter, I summarize these findings and suggest areas for future research. As digital health technology becomes more widely adopted in healthcare delivery, and the risk of disproportionate uptake and use of digital health technology becomes more pronounced for marginalized populations, this dissertation provides a strong scientific framework through which new strategies to reduce disparities, and the digital divide can be developed. These results will inform healthcare systems about innovative and equitable interventions and strategies to reduce inequity in digital health technology uptake and use and promote equity
Designing Amphiphilic Peptoids for Controlling the Formation of Nanostructured Biomimetic Materials
No full textThesis (Ph.D.)--University of Washington, 2025Biomacromolecules such as proteins and peptides achieve remarkable functionality through precise molecular sequences, enabling sophisticated folding and hierarchical assembly. However, their complexity and sensitivity in folding and assembly raise significant challenges for molecular design and limit broader applications. Peptoids, or poly-N-substituted glycines, are a versatile class of peptide mimetics that mimic the sequence tunability while offering unique advantages, including resistance to degradation, environmental stability, and ease of synthesis. With programmable sequences with diverse side-chain chemistries, peptoids enable precise control over molecular interactions and self-assembly pathways, making them ideal for studying and engineering biomimetic nanostructures. These characteristics bridge the gap between biological systems and synthetic materials, providing a robust platform for rational design and predictive synthesis of functional materials. This dissertation focuses on the design and self-assembly of amphiphilic peptoids into nanostructured biomimetic materials, with three fundamental areas of investigation. First, we introduce chiral motifs to guide the assembly of short peptoids into nanohelices, revealing how molecular interactions can be tailored to control structural geometry and handedness. Second, we establish how the assembly of amphiphilic peptoids with anisotropic hydrophobic domains can control morphologies between nanosheets, nanotubes, and nanohelices, developing mechanistic insights into assembly polymorphism. Finally, we expand the complexity of peptoid sequences by introducing multiblock architectures, demonstrating their ability to fold and assemble into hierarchical nanosheets while mimicking protein-like folding behavior. Together, these studies provide a comprehensive framework for understanding peptoid assembly and pave the way for rationally designing bioinspired nanostructured materials with applications like catalysis, sensing, and energy storage
Detecting Breaking Waves and Measuring Bore Speeds in Optical Surf Zone Imagery using Machine Learning
Get PDFThesis (Master's)--University of Washington, 2025A machine learning algorithm is developed to detect breaking waves in optical remote sensing data collected under visually diverse conditions along a kilometer-scale beach in Duck, NC. Bore speeds are estimated from the breaking-wave detections and are compared with theoretical models using surveyed bathymetry. Bathymetry inversion from the derived bore speeds is then explored, revealing low but systematic bias within the surf zone. Despite this limitation, a qualitative analysis of the inverted bathymetry demonstrates that the method captures morphological change over the course of the experiment. This method shows promise as a robust, low-cost approach for measuring wave-breaking patterns and dynamics across large surf zones. The results highlight important considerations for the data resolution, quality, and processing needed to achieve robust measurements of breaking waves using optical remote sensing
Engineering novel nanomaterials through de novo design of hydrophobic scaffold proteins
Get PDFThesis (Ph.D.)--University of Washington, 2025Computationally designed protein nanoparticles leverage de novo designed protein subunits to build large oligomers which can be used for multivalent antigen display and cargo loading in the delivery of vaccines. Advancements in computational methods and design of membrane proteins increases the breadth of materials which can be used as subunit building blocks in these large oligomeric complexes. In this dissertation I describe engineering a protein scaffold with a hydrophobic pore and the rational design of a novel two-component nanoparticle utilizing a de novo designed transmembrane protein. Using machine learning and AI-guided approaches with two-component RPX Docking, AlphaFold 2 and RosettaFold Diffusion, I structurally characterized a novel two component transmembrane nanoparticle and to 4.16 Å by Cryo-EM. This technology marks the first generation of novel hydrophobic nanoparticles that is a step towards generating in vitro hybrid lipid-protein nanomaterials for the display of unique membrane proteins and lipid-conjugated moieties previously inaccessible through current designed nanoparticles
Advances in Open Microfluidics from Fundamental Flow Dynamics to Environmental and Translational Science Applications
Get PDFThesis (Ph.D.)--University of Washington, 2025This dissertation will demonstrate and discuss advances in open-channel microfluidics at the fundamental and translational levels. Chapter 1 outlines new fundamental open-microfluidic tools through via analytical models and comparisons with open channel fluid flow experiments. Chapter 2 will demonstrate enhanced capillary flow through the coupling of homothetic, bifurcating capillary trees and semi-circular paper pads at the extremities to maintain high fluid velocities throughout the channel over an extended period of time. Chapter 3 will discuss and demonstrate the need for a dynamic contact angle (DCA) at high fluid velocities through a survey of current theoretical approaches including multiple hydrodynamic models and the molecular kinetic theory with a comparison to in-lab flow experiments in U-shaped open microfluidic channels. Chapter 4 will present the implementation of trigger valves in open channel configurations allowing for the formation of lateral flow of multiple liquids in parallel for precise fluid addition. Chapter 5 will focus on the use of an open-channel droplet generator that can encapsulate human sperm samples for the use in cryopreservation steps in assisted reproductive technology (ART) workflows
Magneto-Optical Trapping and Control for a Neutral Atom Quantum Computer
Get PDFThesis (Master's)--University of Washington, 2025This thesis presents the design, implementation, and characterization of a Rubidium-87Magneto-Optical Trap (MOT) developed as a part of the foundation of a neutral atom
quantum computing platform. A two-dimensional (2D) MOT and a 2D+ MOT configuration
are realized to generate and deliver a cold atomic beam for future three-dimensional
trapping.
The experimental system integrates laser locking based on saturated absorption spectroscopy,
radio-frequency control of acousto-optic and electro-optic modulators, permanentmagnet
field generation, and a real-time FPGA-based control system. The 2D MOT is
characterized using fluorescence imaging, and the 2D+ atomic beam is characterized by
transversal probe beam spectroscopy. We extract the linewidth and assess Doppler and
power-broadening effects.
The results demonstrate stable generation of a collimated atomic beam and establish
a robust testbed for future integration with optical tweezers and scalable neutral atom
quantum computing architectures