Doctoral Dissertations

Permanent URI for this collection

https://hdl.handle.net/10919/11041

Browse

Now showing 1 - 20 of 17649

The Landscape of Impact: Creating a New Design Framework for Targeted Project Outcomes Through a Practice Based Approach to Research
Bradley, Sharon Eileen (Virginia Tech, 2025-05-16)
The purpose of this research was to explore the potential for design to achieve wide-ranging, positive impacts in targeted and measurable ways. An increasing number of studies indicates that the built environment plays a substantial role in public health and well-being, yet even fundamental human needs such as access to healthy food and safe outdoor spaces are absent in many communities. This research was conducted as a Practice Based Research investigation. It examined a body of work - that of my 30-year practice, Bradley Site Design - and mined it as an empirical resource. Combined with a comprehensive literature review, the investigation formed the basis for a new design methodology and business model that prioritizes the health and wellbeing of people and planet. The examination of project and practice artifacts in the context of established discourse on topics such as impact assessment resulted in a synthesis of applied and theoretical knowledge, revealing new perspectives in design practice. The resulting research product is Design For Impact, a methodology and set of instruments that provide a framework to address localized community needs and generate outcomes that measurably affect human wellbeing, economic stability, and environmental health.
Systems to Transform Interdisciplinary Graduate Education: An Ecological Systems Analysis of STEM Graduate Students' Longitudinal Interdisciplinary Identity-Based Motivation
Webb, Margaret (Virginia Tech, 2025-05-16)
Despite growing recognition that solving complex global challenges requires interdisciplinary approaches, traditional academic structures continue to create significant barriers for STEM graduate students attempting to pursue interdisciplinary work. To address these barriers, this dissertation examines how academic systems influence interdisciplinary identity development and motivation among STEM graduate students through the lenses of Ecological Systems Theory (EST) and Future Possible Selves (FPS). Drawing on longitudinal interviews with graduate students in an Interdisciplinary Disaster Resilience program, this case study reveals complex developmental trajectories, salient microsystems, and system interaction patterns that shape interdisciplinary scholar formation. The research unfolds across three interconnected manuscripts that : 1) identify three patterns of interdisciplinary identity development that challenge linear models; 2) map 12 critical microsystems, spanning past, present, and future, that influence development; and 3) analyze how the core functions of these microsystems act and interact to create supports, barriers, and negotiations in students' development. By integrating EST with FPS, this work demonstrates that interdisciplinary development emerges through interactions among individual aspirations and the entrenched functions of academic microsystems rather than simple acquisition of specific skills or competencies. The findings help explain why sustainable change in interdisciplinary graduate education remains challenging: stable patterns within academic microsystems operate to sustain underlying core functions that actively resist isolated modifications and privilege disciplinary over interdisciplinary development.
Pancreatic Cancer: Oncomicrobes, Electric Fields, and Fluid Flow
Ahmad, Raffae Nazir (Virginia Tech, 2025-05-16)
The pancreatic tumor microenvironment exhibits remarkable complexity, prompting critical investigations into how the microbiota influences tumor progression and how stromal elements impact interstitial fluid dynamics. This dissertation examines dual aspects of this complexity: first, by elucidating the specific contributions of Fusobacterium nucleatum (F. nucleatum) to pancreatic cancer pathogenesis and developing a novel therapeutic approach for eliminating these intracellular bacteria; and second, by analyzing how a stromal-targeted therapy directed at hyaluronic acid modulates interstitial fluid flow within pancreatic tumors using clinical patient data. We demonstrated that F. nucleatum invade both pancreatic cancer cells and normal pancreatic epithelial cells (nPECs). This invasion process was partially mediated by the bacterial adhesin Fap2. nucleatum infection induced a distinct cytokine secretion profile, characterized by elevated IL-8, CXCL1, MIP-3α, and GM-CSF. F. nucleatum invasion promoted migration and proliferation in Panc1 and BxPC3 cancer cell lines. Conditioned media from infected BxPC3 cells stimulated migration in both uninfected BxPC3 and Panc1 cells, suggesting paracrine effects. While F. nucleatum-infected nPECs exhibited a similar cytokine profile, they did not display increased proliferation or self-migration. However, conditioned media from infected nPECs enhanced BxPC3 cancer cell migration, indicating potential cross-talk. Building on these findings, we engineered an electro-antibacterial therapy (EAT) that enhances antibiotic delivery into the intracellular compartment. This approach employs pulsed electric fields to achieve controlled permeabilization of host cell membranes through precise modulation of electric field strength. By combining pulsed electric fields with a standard-of-care antibiotic, we achieved greater than 99% clearance of intracellular F. nucleatum from pancreatic cancer cells. We next examined the broader biophysical features of the tumor microenvironment. We characterized interstitial fluid flow in pancreatic cancer patients, recognizing that desmoplasia creates significant barriers to treatment and influences interstitial fluid pressure. Our analysis of patients treated with PEGPH20, a hyaluronidase enzyme, revealed a transient reduction in velocity magnitudes one day post-treatment, though values generally returned to baseline by the conclusion of the dosing cycle. We observed substantial heterogeneity of velocity magnitudes both within individual tumors and across multiple tumors within the same patient. In one patient with five distinct tumors, we identified variable treatment responses that correlated with tumor size, though velocity magnitude itself did not emerge as a reliable predictor of treatment response.
Exploring Engineering Employment Trends: A Decade-long Deep Dive into Skills and Competences Included in Job Advertisements
Alsharif, Abdulrahman Mohammed (Virginia Tech, 2025-05-15)
My dissertation explores how Natural Language Processing (NLP) can support job advertisement discipline classification to help workforce researchers in analyzing labor market trends and relate it back to higher education. In particular, this study investigates how NLP can be used to identify discipline-specific and education-level skill demands from pre-classified large-scale online job advertisements form Burning Glass Technologies. Although engineering education has made long steps in preparing students with foundational knowledge, employers continue to report a misalignment between the skills students acquire in school and the skills needed in practice. A key challenge in addressing this issue is the effective interpretation of semi-structured labor market data such as online job postings, which contain rich but inconsistently labeled skill information. To address this, I developed an NLP classification system that applies pattern-based text classification and flexible regular expression (regex) matching to identify relevant engineering job postings across Civil (CE), Electrical (EE), and Mechanical (ME) Engineering. The classification framework leverages a dictionary of O*NET job title terms and engineering-specific vocabulary to refine the labeling of jobs originally mapped using Standard Occupational Classification (SOC) codes. To validate the classification accuracy, I evaluated results using confusion matrix metrics (accuracy, precision, recall, F1-score) and performed manual spot-checking of 100 job ads from each discipline. The final classification system achieved high F1-scores across CE (94.2%), EE (91.7%), and ME (93.0%), showing strong alignment with human-judged classifications. This step was essential to ensure accurate discipline-specific labeling for subsequent skill demand analysis. Guided by the SABER-Workforce Development (SABER-WfD) framework, the study then addresses two additional research questions. The second research question examines how skill demands differ by engineering discipline and by degree level (bachelor's, master's, doctoral). Using skill mention proportions and statistical analyses such as ANOVA and Cohen's d, the study reveals that foundational technical skills like Drafting and Engineering Design, CAD, and Microsoft Office tools are dominant across all three disciplines at the bachelor's level. At the graduate level, postings increasingly emphasize management-oriented competencies such as Project Management, Budgeting, and Scheduling, particularly in civil and mechanical engineering. EE showed a higher graduate-level demand for specialized tools like MATLAB, Python, and Simulation. The third research question explores how skill requirements have changed over time from 2010 to 2022. Longitudinal analysis shows a growing emphasis on digital and programming tools (e.g., Python, MATLAB) across all disciplines, especially at the graduate level. Simultaneously, demand for traditional skills such as Drafting, Project Management, and Engineering Design has remained steady or increased, signaling that core engineering competencies remain essential. These time-based trends highlight the dual importance of technical depth and managerial fluency in modern engineering roles. This study demonstrates the potential of NLP-based classification and analysis techniques to extract meaningful trends from complex labor market datasets. In doing so, my dissertation contributes to ongoing discussions about curriculum reform by providing a replicable framework for aligning engineering education with workforce needs. The methodology introduced in this study also offers guidance for researchers and institutional stakeholders aiming to apply NLP in large-scale skill demand analysis, thereby expanding access to labor market insights that support engineering workforce development.
Characterization and Modeling of Deformable Soils for Tire Performance Simulations
Jasoliya, Dhruvin Rakeshbhai (Virginia Tech, 2025-05-15)
The accurate prediction of tire-soil interaction plays a crucial role in the optimization and design of off-road machinery used for agricultural, defense, construction, and mining applications. The tire-road interaction studies focus solely on road roughness and friction between rubber and asphalt while assuming the road surface to be rigid. However, for tire-soil interaction, the deformations occurring in the soil must be accounted for in predicting the tire performance. The modeling of soil presents significant challenges because of its non-linear and complex behavior, which depends on multiple external and internal factors. This thesis focuses on developing a methodology for physics-based modeling of soil for accurate tire performance simulations and its validation. For this study, one cohesive (sandy loam) and one non-cohesive soil (dry sand) are used. The material model parameters of both soils are identified, verified, and validated using a series of laboratory and in-situ test experimental data and numerical simulations. Later, the numerical simulations of tire-soil interaction for both soils are performed using one meshed (Coupled Eulerian-Lagrangian) and one meshless (Smooth Particle Hydrodynamics) method are performed. The results of the numerical simulations are validated with the experimental data obtained at different normal loads (3 kN - 7 kN) and slip ratios (-10% - 40%). Further, the benchmarking of both the numerical methods used is done in terms of relative computational efficiency and accuracy. The cap plasticity material model with strain hardening showed higher accuracy in predicting soil shear strength and failure compared to other material models. For the tire-soil interaction studies, the SPH method overall has a better correlation with the experimental data compared to the CEL method. Meanwhile, the CEL method has higher computational efficiency. The results of the study provide significant insights into the physics of the tire-soil interaction and provide direction for future researchers.
Near-Optimal Sensor Placement for Detection of Poisson Distributed Targets
Kim, Min Gyu (Virginia Tech, 2025-05-15)
In this dissertation, we address the problem of sensor placement for detecting uncertain targets. We model target arrivals using a Poisson process to capture the inherent randomness of event occurrences and emphasize the importance of accounting for uncertainty in the sensor placement strategy. To tackle this, we propose a computationally efficient approximation method based on a lower bound derived from Jensen's inequality. This approach leverages the mean of the uncertain target model to yield a suboptimal yet tractable solution suitable for real-time applications. We evaluate the accuracy of this approximation by quantifying its deviation from the original formulation and providing an upper bound on the approximation error. While the initial framework is formulated in a 1-dimensional spatial domain along a line segment for simplicity, we extend it to a 2-dimensional setting to handle uncertain linear target trajectories using a log-Gaussian Cox line process. Furthermore, we develop an improved closed-form approximation that incorporates both the mean and variance of the target distribution using a second-order Taylor series expansion, offering increased accuracy and a tighter error bound. The effectiveness of our proposed methods is demonstrated using real-world ship traffic data from the Hampton Roads channels in Virginia, USA, obtained from the Office for Coastal Management and the Bureau of Ocean Energy.
At the Intersection of Bistability and Elastic Instability: Switching and Locking Structures using Asymmetric Carbon Fiber Composites
Deshpande, Vishrut Jitendra (Virginia Tech, 2025-05-15)
The next evolution of engineered structures would be to have an on-demand ability to become soft and foldable for packing in compact dimensions. This adaptive capability makes them convenient in transportation such as space structures where packaging of large deployable structures is crucial that help meet ever increasing energy demands of satellites. To render such abilities, smart materials would become necessary – materials that show adaptability to certain stimuli and change one or more characteristic properties. For example, shape memory alloys (SMAs) shrink in length upon heating with increased longitudinal stiffness. Thus, stiffness modulation and morphing ability is a crucial aspect of switchable systems. In our interest, Asymmetric Carbon-Fibre Reinforced Polymer (CFRP) laminates, have shown bistability, i.e., they have two stable equilibria or states that arise due to thermal imbalances during the curing process. This forces the laminate to exhibit two mutually perpendicular characteristic curvatures in two different stable states. The change from one state to the other is termed as a snap-through process. This study for the first time investigates the bistable laminates from a holistic perspective by understanding their quasi-static behaviors in mainly two important scenarios i.e. Out-of-plane and In-plane direction loadings. The authors in their first study uncover the different snap-through mechanics utilizing asymmetric boundary conditions. Three distinct snap-through characteristics are presented — two-step, one-step, and no-snap process — which depends on the load location and boundary conditions. For the second study, in-plane compression testing for bistable laminates reveal two drastically different responses – one very compliant and soft that behaves like a softening non-linear spring, and the other stiff response similar to thin columns which buckle under large loads. This material offers on-demand switching between these two responses by simply snapping their state from one to the other. Through extensive finite element simulation and experimentation, we present effective strategies to enhance their stiff response (buckling load) for improving the stiffness switching ratio. Learning through these behavioral characteristics of bistable laminates, a novel concept is implemented for morphing structures. A 'locking' feature is introduced by harnessing characteristic curvatures of these bistable laminates and strategically implementing them in morphing structures. We take inspiration from various origami folding patterns and incorporate a waterbomb geometry in these bistable laminates. This helps in changing the load-bearing capabilities of the bistable laminate, by switching from a very soft foldable state to a lockable stiff state. We present a case study on two origami designs, namely Kresling and Yoshimura, where using this bistability property delivers massively reconfigurable structures that show meta-stable load-bearing states.
Exploring the relationships among impulsivity, interpersonal difficulties, and social risk-taking in borderline personality disorder: behavioural influences and neural correlates
Palissery, Gates Krystal (Virginia Tech, 2025-05-15)
Impulsivity and difficulty maintaining interpersonal relationships (interpersonal difficulties) are two symptoms characteristic of borderline personality disorder (BPD). Past work has shown that these symptoms may have differential effects on risk-taking behavior: impulsivity has been associated with increased risk-taking, while interpersonal difficulties has been associated with decreased trust in social contexts, which can be construed as decreased risk-taking behavior. More work is needed to understand how these symptoms are related to monetary risk-taking in social contexts. The goal of this dissertation is to better understand the relationships symptoms of impulsivity and interpersonal difficulties have with monetary risk-taking in a social context in individuals displaying a range of BPD features (Study 1) and individuals with BPD (Study 2). This dissertation further seeks to elucidate the neural correlates of social risk-taking as they pertain to these sets of symptoms in individuals with BPD (Study 3). Study 1 finds that individuals with increased symptoms of impulsivity are less sensitive to the difference between two gambles' risks, and individuals with increased symptoms of interpersonal difficulties are less sensitive to the difference between two gambles' risks only in a social context. Study 2 finds that individuals with increased symptoms of impulsivity are less sensitive to the difference between two gambles' risks, and individuals with increased symptoms of interpersonal difficulties are less sensitive to the difference between two gambles' risks regardless of context. Study 3 finds that individuals with BPD show decreased anterior insula response to the difference in risk between the gamble they select and the unchosen gamble, though it did not find a relationship between insula response and BPD symptoms. Together, these studies show that individuals with BPD respond differently to monetary risk-taking than individuals without BPD. This work suggests symptoms of impulsivity and interpersonal difficulties may be new targets for behavioral interventions to treat BPD.
Processing Approaches for Maintaining Multifunctionality in Advanced Thermoplastic Composites
Anderson, Justin (Virginia Tech, 2025-05-15)
Multifunctional composites typically consist of a functional additive imparting multiple functionalities to a polymer matrix. The production of these composites is not trivial, as the functionality of the particles must be retained during production, and the polymer matrix and particle interactions can often reduce the effectiveness of the final product. When fundamental principles, including thermodynamic and hydrodynamic considerations, are utilized in the production of these materials, it is possible to create highly effective multifunctional composites with tremendous application potential. Common production methods for multifunctional composites include extrusion, solution casting, molding, and spinning techniques. Most of these techniques are used to create materials with a homogeneous microstructure, as thorough mixing is needed for the matrix material to successfully accept the functional particle. In solvent casting, this particle dispersion in the matrix is often controlled through a solvent exchange step, as particles must frequently be transferred from a suspension or colloid that is not miscible with the desired polymer matrix to a solvent that is. Another method to create well-dispersed particle composites is by creating a layered structure of two separate matrices. By embedding a functional particle into only one of these matrices, the particle density can effectively be controlled, and therefore the surface area to volume ratio is increased. This type of layered system is often produced through melt pressing or extrusion methods and requires a keen understanding of how to control the organization of a particle in the presence of two dissimilar matrices. Chapters 3 and 4 will jointly address the issue of transferring a functional particle from a colloid into a well dispersed thermoplastic matrix while also introducing a novel use for these materials through solution casting. The material studied consists of a modified cellulose nanocrystal (CNC) with additional carboxylated "hairs" called electrosterically stabilized nanocellulose crystals (ENCC). By embedding these particles in a thermoplastic polyurethane (TPU) matrix, a stimulated optical response is achievable through hydration of the bulk film. This phenomenon is reliant on the dispersion of the ENCC particles, as particle agglomeration will result in non-uniform opacity. Chapter 3 will address the issue with the solvent exchange of a charged tertiary mixture. ENCCs are highly hydrophilic while generally being stored in a colloid. This leads to difficulties when attempting to solvent exchange the colloid with a solvent that is miscible with both the ENCC and TPU. By modifying the existing solvent exchange process for CNC-TPU composites to account for the additional hydrophilic hairs of the ENCC, we were able to successfully transfer the ENCC from a water to DMF mixture and cast ENCC-TPU films of varying concentrations. Chapter 4 will further develop the understanding of these ENCC-TPU composites and the effects of ENCC particle dispersion on their mechanical and optical properties. The effects of particle dispersion on the mechanical and optical properties of ENCC-TPU films were probed through dynamic mechanical analysis (DMA) and ultraviolet-visible spectroscopy (UV-VIS) respectively. Chapter 5 will address the issue of producing a multilayered filament while containing a functional particle to one of the layered matrices, and the effects of multiple processing methods on the particle functionality. Melt pressing and multilayer extrusion were used to produce antiviral particle filled composites layered with a neat matrix. When the antiviral particles at the surface of the functional composite layer are rendered ineffective, the top layer of the multilayered system can be delaminated to expose a surface with fresh functional particles. The functional particles used in this study, cuprous oxide (Cu2O) and a polymer-based antiviral particle (AV1), are well studied antiviral agents. The antiviral performance can be measured pre and post extrusion giving insight into the effects of particle viability during processing. Through careful selection of the matrix materials, we were able to successfully produce multilayered systems consisting of alternating layers of neat polypropylene (PP) and a AV1-LDPE composite. Melt pressed samples showed minimal particle diffusion across the PP-LDPE interface, while the AV1 antiviral efficacy was only slightly reduced from 99.9% to 86.0% viability against COVID-19. We also produced multilayered filaments of the same layered system while also retaining uniform layering and minimal particle migration.
Finite-Difference and Analytic-Gradient Approaches for Simulating Vibrational Circular Dichroism Using Second-Order Møller-Plesset Perturbation Theory and Configuration Interaction
Shumberger, Brendan Michael (Virginia Tech, 2025-05-14)
Vibrational circular dichroism (VCD) is defined as the differential absorption of left- and right-circularly polarized light in the infrared region of the electromagnetic spectrum. Application of this spectroscopy is primarily directed towards the elucidation of molecular absolute configuration. As a result of the complex relationships involved in light-matter interactions, theoretical simulation is required to interpret experimental results. In this work, we focus on improving the accuracy and efficiency of simulating VCD spectra. First, we discuss the effects of the choice of basis set on two chiroptical properties including VCD and Raman optical activity (ROA) with a particular emphasis on property-oriented basis sets. Next, we introduce a finite-difference scheme for computing the atomic axial tensor (AAT), a required quantity for VCD simulation, for the second-order Møller-Plesset perturbation (MP2) theory and configuration interaction with double excitations (CID) electronic structure methods. Finally, we formulate an analytic implementation of the MP2 and configuration interaction including single and double excitations (CISD) AATs.
Resistance Exercise and Glycemic Health: Investigating the Impact of Resistance Exercise in Healthy Individuals and Those with Impaired Glucose Homeostasis
Reynolds, Jake Colton (Virginia Tech, 2025-05-14)
Resistance exercise (RE) is a powerful exercise modality used across the lifespan to promote healthy skeletal muscle mass. While most engage in RE to improve strength, size, and physical functionality, there are metabolic advantages for maintaining healthy levels of skeletal muscle because it is the primary storage site of glucose. Understanding this, RE may be important, especially in individuals with impaired glucose homeostasis. While we have clear evidence that RE improves glycemic outcomes in individuals with impaired glucose homeostasis it is unclear to what degree shifts in lean mass (LM) moderate these changes. Understanding how shifts in LM as well as RE prescription features leading to improved glycemic outcomes may better inform health practitioners working with this population. In contrast, in younger, athletic individuals, RE is commonly used in conjunction with a caloric surplus, to intentionally gain body mass to improve sports performance, improve aesthetics, or improve occupational performance in populations such as military personnel. However, the metabolic implications of this practice are largely unresearched. The objectives of this dissertation are to determine: (1) the effects of intentional weight gain with RE on fasting blood glucose (FBG), 24-hr mean glucose, and glycemic variability using continuous glucose monitoring (CGM) in younger athletic individuals and (2) the effects of RE-induced lean mass in populations at risk of or with type 2 diabetes To assess how CGM has been utilized in conjunction with RE in normoglycemic individuals a narrative review was conducted. The search located 3 clinical trials and 2 observational trials. Based on the observed literature there is currently not a justification for the use of CGM in conjunction with CGM to enhance RE outcomes or glycemic health. In addition, the available studies evaluated acute glycemic responses to RE rather than assessing chronic adaptations after a period of training. This presents a gap in literature that warrants future research. In conjunction with a larger trial investigating weight gain in athletic individuals we determined the effects of intentional weight gain on glycemic health utilizing CGM-derived indices of glycemic health and fasting blood glucose (FBG). Thirty-two athletic individuals were randomized to a peanut containing snack group (PNT) or a higher carbohydrate containing snack group (CHO), each designed to provide participants with an additional 500 kcal/day. Participants also engaged in a whole-body hypertrophy-focused RE protocol 3 days/week for 10 weeks. CGM devices were worn for a 3-day wear period before the intervention and during week 10. Dual-energy X-ray absorptiometry was used to measure body composition pre- and post- intervention. Linear mixed effects models were used to evaluate how changes in body composition affect glycemia. The final analysis included 26 individuals (46% female) aged 24 ± 6 years, with a BMI of 23 ± 3 kg/m². Total weight gain was different between snack groups (PNT: 1.49 ± 1.18 kg, CHO: 2.88 ± 1.18 kg; p= 0.006) and fat mass (FM) gain was higher in CHO (0.98 ± 0.82 kg) compared to PNT (0.50 ± 0.40 kg) (p=0.016). There were no group differences in LM gain (PNT: 1.26 ± 1.06 kg, CHO: 1.90 ± 1.01kg; p= 0.133). There were no group differences in CGM-derived indices of glycemia or FBG. FBG tended to increase with FMI (β=2.32; 95%CI: 0.06, 4.58; p= 0.045). Lean mass index (LMI) was associated with increased time in range (TIR) (β= 2.46; 95%CI: 0.46, 4.45; p= 0.018) as well as mean glucose (meanG) (β= 2.52; 95%CI: 1.32, 3.71; p< 0.001). MeanG had a three-way interaction for group, LMI, and FMI (β= 1.68; 95%CI: 0.18, 3.18; p= 0.031) and LMI, FMI, time (β= 1.35; 95%CI: 0.01, 2.70; p= 0.049). CGM-derived indices of glycemic health did not change after 10 weeks of intentional weight gain via resistance exercise and energy surplus in athletic, normoglycemic individuals. LMI and FMI are associated with shifts in CGM-derived indices of glycemic health and may be linked in their influence on certain measures such as meanG. Additional research is needed to expand this potential relationship to determine how body composition may influence CGM-derived measures of glycemic health. To determine if RE-induced gains in LM effects glucose homeostasis in individuals with impaired glycemic control we conducted a systematic review. We observed a mean gain in LM was ≥ 2kg was associated with the largest decrease in HBA1c. Additionally, the RE prescription that appeared to be most effective in improving HBA1c utilized ≥ 9 sets per muscle group/exercise/week of ~60%1RM or more, for ≥ 8 repetitions, and a duration of 12-26 weeks.
Bridging Multimodal Learning and Planning for Intelligent Task Assistance
Tabassum, Afrina (Virginia Tech, 2025-05-14)
Task-assistance systems provide adaptive, multimodal guidance for complex, step-based activities such as cooking and DIY projects. A central challenge lies in enabling these systems to interpret real-world scenarios—understanding user intent from verbal, visual, or textual cues and generating coherent, multimodal instructions enriched with relevant visual. To tackle this, modern systems leverage advanced machine learning techniques, from representation learning that processes information from diverse modalities (e.g., text, images, audio) to procedural planning which provides dynamic, context-driven guidance, enabling systems to provide precise, real-time assistance tailored to user needs. This work addresses core challenges in representation learning and multimodal planning through three key contributions. First, we introduce a modality-agnostic contrastive learning framework that optimizes negative sample selection by jointly balancing anchor similarity, influence and diversity, improving generalization across vision, language, and graph tasks. Second, we propose a tuning strategy for masked audio models that leverages unsupervised audio mixtures to enhance adaptation to downstream tasks with less labeled data, such as few-shot learning. Third, we present a zero-shot framework for generating multimodal procedural plans with explicit object-state consistency, paired with two novel evaluation metrics and an evaluation task to assess planning accuracy, cross-modal alignment and temporal coherence. These contributions are integrated into a context-aware multimodal task assistant, empirically validated through real-world user studies. Our work establishes a foundation for more robust, adaptable, and user-centric task-assistance systems, bridging critical gaps in multimodal understanding and guidance.
Neurological Consequences of Viral Encephalitis: Behavioral Deficits, Neuronal Restructuring, and Therapeutic Interventions
VanderGiessen, Morgen (Virginia Tech, 2025-05-14)
Viral infections may lead to persistent neurological symptoms that can mimic or potentially trigger the onset of neurodegenerative disorders linked with aging. Neuronal degeneration, a hallmark of neurodegenerative diseases and viral neuropathologies, involves the progressive loss of neuronal structure and function, often leading to memory dysfunction, motor impairments, and chronic inflammation. We study the cognitive effects of Venezuelan equine encephalitis virus (VEEV), a neuroinvasive alphavirus that causes severe neurological symptoms, including seizures and encephalitis, with no available treatments or vaccines. Despite its significant public health implications, including its potential as a biological weapon, VEEV remains understudied, particularly regarding its chronic effects on memory and neuronal function. Therefore, this work bridges the gap to identify critical features of neuron loss and neuron restructuring that lead to neurobehavioral deficits and explore two avenues of broad-spectrum antiviral and neuroprotective therapeutics. First, we investigate the acute and chronic neurological consequences of VEEV infection using a murine model, focusing on behavioral, neuropathological, and transcriptomic changes. Mice displayed decreased anxiety-like behavior, decreased recognition memory and altered neuromuscular functions during chronic disease. Single-cell transcriptomic analysis showed that innate immune and inflammatory responses were activated at an acute time post-infection and neurological signaling was dampened during chronic disease. VEEV infection resulted in chronic activation of microglia and astrocytes and chronic neuron loss in the hippocampus, which correlated with the altered transcriptomic profile observed in the hippocampus. A key therapeutic focus of this study is the tumor suppressor protein p53, a pivotal regulator of apoptosis and cellular stress responses. Activating p53 using the small molecule inhibitor NVP-CGM097 resulted in significant inhibition of VEEV infectious titers. Secondly, we investigated the therapeutic potential of Pifithrin-μ (PFT-μ), an inhibitor of the autophagy-associated chaperone HSP70. PFT-μ treatment resulted in a significant reduction in viral replication for VEEV and related alphaviruses. In mice, PFT-μ treatment reduced weight loss, neurological symptoms, and alternations in anxiety-like behaviors and neuromuscular functions. Together, these insights bridge cell biology, virology, and neuropathology and offers innovative strategies to understand and combat alphavirus-induced neuronal damage. It provides a valuable framework for developing antiviral therapies and neuroprotective interventions, with broader implications for understanding viral impacts on cellular function and neurodegeneration.
Mathematical models of immune responses during external challenges and autoimmunity
Murphy, Quiyana Monet (Virginia Tech, 2025-05-13)
Characterizing the mechanisms of the immune system and its response to infection and autoimmunity is crucial for understanding and predicting disease outcomes and evaluating potential interventions at multiple scales. While data provide a snapshot of certain biological phenomena, they cannot capture the underlying dynamics. Despite technological advancements, data limitations—arising from ethical, technical, and financial constraints—continue to hinder the precise quantification of key biological processes involved in disease progression and transmission at both individual and population levels. Mathematical models have been used with data at multiple scales to investigate the underlying complex systems of the body's natural molecular, cellular, and systemic processes, its response to external challenges or immunodeficiencies, and subsequent impacts on disease, treatment, and transmission outcomes. This dissertation uses mathematical modeling, mathematical analysis, and parameter estimation tools to uncover mechanisms of immune responses and immune system dysfunction during autoimmunity and viral infections. The immunological and viral dynamic models were validated against longitudinal virus or immune cell and marker data. The immune response is the body's mechanism for protecting itself against foreign pathogens (e.g., viruses, bacteria, fungi, toxins) or substances it deems harmful. It consists of the innate (non-specific) immune response and the adaptive (specialized) immune response. The innate immune response serves as one of the body's first lines of defense, responding rapidly and uniformly to all foreign substances. This pathogen-induced innate immune response operates through inflammatory immune mechanisms and the removal of foreign particles by immune cells. In the case of SARS-CoV-2, virus-induced persistent inflammation and tissue damage are associated with increased COVID-19 severity and disease progression. To better understand the mechanisms underlying persistent inflammation in severe COVID-19 cases, we developed a mathematical model of the innate immune response following SARS-CoV-2 infection. After fitting the model to immune cell and immune marker data from COVID-19 patients, we estimated key parameters for both mild and severe clinical cases. Analytical, bifurcation, and numerical techniques were used to investigate how changes in immune function affect long-term immune dynamics and to identify potential mechanisms needed for immune resolution. If the innate immune response fails to eliminate the pathogen, the adaptive immune response takes over. Although it is slow to activate, the adaptive immune response employs specialized immune cells and antibodies that specifically target and eliminate foreign pathogens while generating memory cells that allow for a faster response upon reinfection. Immune memory can also be induced through vaccination, as demonstrated by the global vaccination efforts during the COVID-19 pandemic. However, we observed that vaccine-induced immunity to SARS-CoV-2 waned over time as mutations in the virus emerged, necessitating booster vaccinations to maintain protection. Accurately predicting the strength and durability of the adaptive immune response requires understanding diverse immune profiles. In the case of SARS-CoV-2, immune profile heterogeneity arises from prior infection with multiple variants, the severity of previous infections, and the timing and order of natural infection and vaccination. We developed mathematical models of antibody responses following vaccination against SARS-CoV-2, fitting them to longitudinal antibody data to determine the long-term dynamics and composition of the antibody-mediated immune response in individuals with varying immune landscapes. When functioning correctly, the adaptive immune response can last weeks, months, or even years, depending on the context. However, in some cases, the adaptive immune response fails to distinguish between foreign and self-antigens, leading to an immune response against the body's own tissues. One example is systemic lupus erythematosus (SLE), a chronic autoimmune disease with no single known cause, though evidence suggests that dysregulated immune responses—driven by a combination of genetic predisposition and environmental factors—may contribute to its development. One emerging area of interest in immunotherapy is centered around harnessing the anti-inflammatory properties of Aryl hydrocarbon receptor (AhR), a flexible ligand-activated transcription factor that acts as an environmental sensor, activation. Using mathematical modelling and data from an animal study, we developed a framework for identifying cellular and molecular factors that contribute to physiological outcomes observed in Lupus. We developed a novel model to describe dynamics of immunosuppressive and follicular T cell phenotypes and predicted the T cell balance over time. Lastly, we developed a multiscale model to describe disease dynamics in an emerging zoonotic disease, Usutu virus, which spreads between birds and mosquitoes with occasional spillover humans causing neurological disease. The multiscale model was fit to data at three biological scales: infectious Usutu virus titers in canaries, bird-to-mosquito transmission data, as well as reported susceptible and infected birds. The model was used to predict disease dynamics at multiple scales. Since during data fitting, the model and type of data being used has significant effect on reliability of parameter estimates and predicted disease dynamics, we conduct identifiability analyses to determine the reliability of parameters estimated from our model.
The Role of Food Retailers for Integration of Nutrition Security and Planetary Health Promotion in the United States
DeNunzio, Maria Nicole (Virginia Tech, 2025-05-13)
In the United States (U.S.), nutrition security is defined as food security and diet quality across social and economic segments of the population. Achieving nutrition security for all requires a healthy planet that can withstand the shocks and stressors of ahistoric weather patterns. Healthy food retail initiatives in the U.S. promote nutrition security through policy, systems, and environmental changes, but under-emphasize planetary health promotion. The purpose of this dissertation was to explore the role of food retailers in the U.S. to integrate nutrition security and planetary health promotion. Study one examined the applicability of an existing corporate food retail benchmarking tool, the Business Impact Assessment-Obesity and population level nutrition (BIA-Obesity), for monitoring food retailer actions to advance the National Strategy on Hunger, Nutrition, and Health recommendations for nutrition security. Existing BIA-Obesity indicators were broadly applicable for National Strategy recommendations. Additional indicators for local foods, fairness in resource distribution, food waste reduction, and digital food environments would improve applicability of the BIA-Obesity to monitor food retailer actions towards National Strategy recommendations. Study two identified corporate Supplemental Nutrition Assistance Program (SNAP)-authorized food retailer public commitments to environmental sustainability and categorized commitments by domain of environmental sustainability, using relevant Sustainable Development Goals (n=9). Content analysis of commitments showed that 31 of 48 included SNAP-authorized food retailers had commitments across environmental sustainability domains. Results can inform accountability evaluations, partnerships, and policy action. Study three explored the perceptions of 12 independent food retailers in Virginia about planetary health promotion through semi-structured interviews, with questions and coding informed by the inner and outer settings of the Consolidated Framework for Implementation Research. Food retailers were willing to play a role in planetary health promotion and identified cost and customer satisfaction as key determinants of the scope of their role. Planetary health promotion practices of interest varied by store format and community context. Corporate and independent food retailers could support planetary health if support and incentive systems align planetary health promotion with profit potential and customer satisfaction. Healthy food retail researchers and practitioners can use these results to inform expanded nutrition security programming that includes planetary health promotion.
Flexible Control and Stability Paradigms for the Inverter-Dominated Power System
Venkataramanan, Ashwin (Virginia Tech, 2025-05-13)
The addition of inverter-based resources (IBR), distributed energy resources (DER), and retirement of synchronous generation has resulted in significant changes to the transient characteristics of the bulk power system (BPS). This dissertation proposes new control and stability paradigms that leverage the flexibility offered IBRs and DERs in shaping the power system transient response and creating a stable power system. This dissertation leverages the controllable nature of IBRs to propose a flexible control paradigm where the primary control performance of existing inverters is modified to meet the transient performance needs of an evolving power system. A model-free and black-box control strategy is proposed to shape the transient response of existing IBRs without access to their internal parameters. A reinforcement learning (RL)--based algorithm is proposed to utilize local measurements to enable black-box control of IBRs in transient timeframes. Further, Lyapunov functions are learned from measurements and integrated to the RL-based strategy for stable control. Additionally, the dissertation proposes a flexible stability paradigm that explores the stabilizing aspects of a distribution system with high DER penetration, rather than considering the distribution system in the traditional context as a simple unidirectional power delivery system. The characteristic aspects of transient stability in a distribution system with high penetration of DERs are analyzed along with the role of fast DERs in impacting BPS stability. The dissertation further proposes a new voltage sensitivity-based screening metric to evaluate relative stability of distribution systems with high DER penetration.
Rhythm and Roots: A Black Feminist Exploration of Culturally Relevant Pedagogy in Agricultural Education
Spencer, Kendrick LeRoy (Virginia Tech, 2025-05-13)
The American education system is characterized by significant diversity, with students representing a wide range of ethnic, cultural, linguistic, and socio-economic backgrounds. To effectively support all learners, teachers require the necessary tools and resources to implement inclusive and equitable teaching strategies. Educational scholars have utilized asset-based pedagogies to enhance learning and academic achievement, particularly for students from marginalized communities. One such pedagogical approach is Culturally Relevant Pedagogy (CRP). While the broader field of education has embraced and implemented CRP, school-based agricultural education (SBAE) teachers have primarily been exposed to multicultural education without deeper engagement in culturally responsive teaching methods. Given that agricultural education classrooms are as diverse as general education settings, SBAE teachers must be equipped with effective instructional strategies to support all students. This study employs a three-part, multi-modal research design informed by Black Feminist Thought to examine how agricultural teaching practices align with Culturally Relevant Pedagogy. The first study, a national survey of 514 agricultural educators, revealed that teachers highly valued instructional practices related to teaching and reflection, as well as collaboration with community members. However, they placed the least value on enacting social justice and engaging in diversity-related professional development. The second study, a focus group with Black women agricultural science teachers, uncovered their experiences with racism and microaggressions from colleagues, their commitment to political clarity in protecting and supporting their students, their deep care for student success, and their high academic and post-secondary expectations. The final study analyzed the teaching practices of 26 agricultural educators across the country. Findings indicated that while teachers recognized that success varies for each student, their definitions of success differed. They maintained high expectations for students, fostered warm and welcoming classroom environments, and valued their local communities as assets to agricultural education. Participants believed in the efficacy of hands-on learning and valued the opportunities provided through the National FFA Organization (FFA). However, they did not exhibit critical consciousness related to their teaching strategies.
Conversion of Switchgrass into Functional Carbon Materials and Food Emulsifiers
Li, Yilin (Virginia Tech, 2025-05-13)
Effects of dietary feed additives on the performance and gastrointestinal health of broilers exposed with coccidial spores and varying litter conditions
Fritzlen, Cooper James (Virginia Tech, 2025-05-13)
Coccidiosis is a prevalent parasitic disease in broiler chickens, caused by protozoa of the genus Eimeria. Birds become infected by ingesting sporulated oocysts from contaminated litter, resulting in intestinal damage, impaired nutrient absorption, and reduced growth performance, which contribute to economic losses for poultry producers. Conventional control methods, including vaccines, ionophores, and other anticoccidials, have limitations, prompting the need for alternative strategies. This dissertation investigates the effectiveness of dietary feed additives, essential oils (garlic and cinnamon), betaine, butyrate, and benzoic acid, on broilers raised under coccidial challenges and varying litter conditions such as differences in moisture content and reuse of litter, focusing on their potential to improve performance, gastrointestinal health, and sustainability while reducing antimicrobial reliance. The second chapter provides an in-depth literature review on coccidiosis, alternative feed additives, and factors influencing litter quality. In the third chapter, the combination of essential oils and betaine was effective in mitigating the effects of coccidiosis. Broilers supplemented with 500 ppm of essential oils and 250 ppm of betaine and raised on used litter seeded with coccidia increased BWG and reduced FCR over multiple time periods (P ≤ 0.05). Oocyst counts were reduced in treated birds compared to the PC, validating the efficacy of this treatment in mitigating the adverse effects of coccidiosis. The fourth chapter evaluated butyrate supplementation in broilers raised on used litter without antibiotics, examining its effects on growth performance, nutrient digestibility, and intestinal health. The supplementation of butyrate in the broiler diet resulted in improved performance over the starter phase of broiler production. Broilers fed 500 ppm butyrate from days 0–8 exhibited a 7% improvement in BWG and a 5% reduction in FCR compared to the PC (P ≤ 0.05). However, these advantages diminished by day 42. Additionally, butyrate improved apparent ileal digestibility of dry matter and energy during the starter phase (P ≤ 0.05) but did not significantly reduce oocyst shedding. Benzoic acid was investigated in pure (PBA) and enteric released (EBA) forms in the fifth chapter. Over the 0–42-day period, EBA supplementation (330 ppm) improved BWG by 5.1% and decreased FCR by 4.6% compared to the control group, which consisted of broilers fed a basal diet without benzoic acid supplementation (P ≤ 0.05). Additionally, enteric-released benzoic acid led to significant improvements in intestinal morphology, specifically reducing crypt depth and increasing the villus height-to-crypt depth ratio in the jejunum and ileum. In the sixth chapter, the combination of betaine and butyrate (B+B) was evaluated under varying litter moisture conditions. B+B supplementation increased energy (4.0%) and dry matter digestibility (4.5%) on day 17 (P ≤ 0.05) but did not improve performance over a 0–42-day broiler grows out period. The increased litter moisture generated by the addition of 2,000 mL of water/day (over the first 21 d) resulted in decreased broiler BWG over the starter phase in comparison to the broilers raised on litter without water addition (P ≤ 0.05). The seventh chapter builds a tool to allow the use of Near-Infrared (NIR) spectroscopy for real-time litter moisture monitoring. To achieve this, 1,455 litter samples across five trials were modeled for NIR spectrum and litter dry matter. Calibration models exhibited high predictive accuracy, characterized by improved R-squared values and reduced standard errors of prediction. These findings highlight the promise of NIR spectroscopy as a tool for precision litter management, particularly with continued advancements and updates to on-farm technology. Overall, this dissertation defines the potential of feed additives in broiler production and the importance of managing environmental conditions, particularly litter quality to maximize performance and efficiency. These findings provide potential and practical insights into reducing reliance on antibiotics while maintaining broiler performance and health, aligning with industry demands for sustainable broiler production.
Neuromuscular Dysfunction as a Novel Indicator for Alzheimer's Disease and Response to Intervention in the 5xFAD Model
Brisendine, Matthew Henry (Virginia Tech, 2025-05-13)
Alzheimer's disease (AD) develops along a continuum that spans years to decades before cognitive decline and clinical diagnosis. Preclinical AD is often associated with decreased muscle function and mitochondrial respiration, but the cause of these peripheral phenotypes in a brain disease remains unclear. Exercise promotes muscle, mitochondrial, and cognitive health, and is proposed as a potential therapeutic for AD. This study investigates skeletal muscles adaptation to exercise in an AD-like context using 5xFAD mice, an AD model developing early pathology and cognitive impairments around 6 months of age. We examined in vivo neuromuscular function in both muscle and the sciatic nerve, and exercise adaptations (mitochondrial respiration and RNA sequencing) before overt cognitive impairment. We found that 5xFAD mice develop neuromuscular dysfunction as early as 4 months of age, characterized by impaired nerve- stimulated muscle torque production and sciatic nerve compound action potential. Additionally, skeletal muscle in 5xFAD mice showed sex-dependent altered adaptive responses to exercise training without cognitive impairment. Given these findings, we hypothesized that voluntary wheel running or an acetylcholinesterase inhibitor donepezil treatment, started before neuromuscular decline, would delay neuromuscular impairment in 5xFAD mice. Using 3-month-old 5xFAD and wild type (WT) littermates, we provided voluntary wheel access for 4 weeks and assessed exercise capacity, tibial nerve- stimulated plantar flexion torque, and sciatic nerve compound action potential at 4 months. Additionally, we investigated markers of AD-like pathology, such as amyloid- beta and neurofilament light chain. In a separate cohort, we treated 3-month-old 5xFAD mice with donepezil or placebo daily for 4 weeks and assessed similar outcomes. Our data show that both interventions delay neuromuscular impairment from 3 to 4 months but do not improve muscle-torque production as seen in exercise-trained WT mice. Neither intervention altered markers of AD-like pathology. Declines in peripheral systems, such as skeletal muscle, may be preclinical identifiers for AD, and interventions like exercise or acetylcholinesterase inhibitors may delay their manifestation.

Browse

Recent Submissions