Doctoral Dissertations

Permanent URI for this collection

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

Browse

Now showing 1 - 20 of 18365

Characterization of canonical and noncanonical nickel metallochaperones in Methanococcus maripaludis
Dinh, Thuc Anh (Virginia Tech, 2026-06-01)
Nickel is an essential trace element for many microorganisms, serving as a catalytic cofactor for enzymes involved in energy metabolism, carbon cycling, and microbial physiology. At the same time, excess intracellular nickel is toxic, requiring organisms to maintain highly regulated nickel homeostasis systems that balance nickel acquisition with controlled intracellular trafficking. Methanogenic archaea (methanogens) rely on multiple nickel-dependent enzymes that are central to methanogenesis, including [NiFe] hydrogenases. The nickel insertion step during [NiFe] hydrogenase assembly involves the nickel metallochaperones HypA and HypB. HypA is a small nickel carrier that contains an N-terminal nickel binding domain and a C-terminal structural zinc-binding site. HypB belongs to the P-loop family of GTPase with a nickel-binding site at the G-domain. In this work, we characterized the nickel metallochaperones HypA and HypB from the hydrogenotrophic methanogen Methanococcus maripaludis. MmpHypA binds either zinc or a mononuclear iron at its C-terminal metal-binding site, the latter representing an unusual metal occupancy not previously observed for HypA proteins. MmpHypB contains the canonical nickel-binding site at the G-domain that stimulates GTP hydrolysis and promotes the formation of HypA-HypB complexes. Zinc-bound MmpHypA is optimized for nickel acquisition of MmpHypB, and the nucleotide state of MmpHypB modulates the oligomeric state of HypA-HypB complexes, supporting a GTPase-driven nickel delivery mechanism. In addition to the canonical HypB protein, methanogens encode a second HypB homolog, HypB2, whose exact function remains unknown. Biochemical characterization of HypB2 proteins revealed that they form a stable complex with proteins encoded by neighboring genes. HypB2 exists predominantly as dimers and the dimeric form is stabilized by strand-swapping at the extended C-terminal region. This region also putatively coordinates a Fe-S cluster, a feature not previously associated with HypB proteins. Deletion of hypB2 does not impair growth under nickel-limiting conditions and HypB2 did not form complex(es) with HypA, suggesting that HypB2 represents a distinct member of the P-loop GTPase family that may not function as a canonical nickel metallochaperone. Instead, AP-MS experiments hint toward a role of HypB2 and its associated complex in Fe-S cofactor trafficking. Together, these findings expand current understanding of nickel trafficking in methanogens and reveal previously unrecognized diversity in the nickel metallochaperone family as well as identify a protein complex that potentially involves in Fe-S cofactor trafficking in methanogens.
Evaluating the Relationship Between Equine Pain and Behavior
Thompson, Rebecca Anne (Virginia Tech, 2026-06-01)
Pain is physical and emotional discomfort caused by a variety of diseases and injuries. While the exact prevalence of pain in horses is unknown, it is likely a common ailment. Additionally, previous research has determined that owners frequently do not recognize pain in horses. This dissertation examined both assessment of pain and a disease that might cause pain. First, I compared responses to two commonly used pain assessment methods: manual palpation and pressure algometry, and used a cluster method to group the responses. Responses were behaviors such as a muscle twitch, tail swish or step. Manual palpation uses a person's hands to touch various areas on a horse to see if the horse responds with a behavior that suggests pain. Pressure algometry uses a digital device that gives a number called the mechanical nociceptive threshold, which is the amount of pressure that was applied when the horse responds with a behavior. There were 46 manual palpation points and 15 pressure algometer points. These points were divided into nine anatomical regions: head, neck, withers, shoulder, front leg, girth, back, hip, and back leg. From the first study, I determined that behavior responses by anatomical region were similar for both manual palpation and pressure algometry testing and therefore only one of these tests was needed if only examining behavior responses. A cluster analysis grouped both the pressure algometer behavior responses and mechanical nociceptive threshold into clusters that were based on anatomical regions. However, a cluster analysis did not create anatomical groups for the manual palpation test. For the second study, I used three pain assessment methods: Equinosis Lameness Locator, manual palpation test, and pressure algometer test to evaluate horses with and without outer surface protein F antibodies to Borrelia burgdorferi. The Equinosis Lameness Locator is a three-sensor system that provides three numerical outputs for any movement asymmetries in millimeters. For this study, I found that horses with B. burgdorferi antibodies had decreased head movement responses on the caudal neck and a higher muscle twitch response count during the pressure algometer exam. However, these horses also had a decreased Equinosis Diff Max Pelvis number meaning that horses with antibodies were moving more symmetrically than horses without antibodies. Additionally, while not related to the tests used to evaluate pain, the overall seroprevalence of horses has increased in the southwest Virginia area. For the third study, I used a sound sensitivity test to determine if horses in varying lameness groups responded differently to repeated sounds as previous research in dogs has found that dogs with orthopedic pain may develop sound sensitivities. In this study, 70% of the horses had decreased behavior reactions when comparing the day one behavior after the sound to the day five behavior after the sound. This means that most horses that are exposed to repeated sounds will habituate to the sound. However, five horses (14%) had increasing behavior between day one and day five. These five horses were in the moderate and high lameness group and there were no horses in the low lameness group that had an increased behavior reaction between day one and day five. Overall, results from these studies help improve our understanding of pain assessments and behaviors; and also, how to apply these assessments to a disease.
Design and SDR Prototyping of Online Real-Time Machine Learning for NextG MIMO and ISAC
Liang, Yibin (Virginia Tech, 2026-06-01)
Signals and market valuation in tourism: Strategic conditions and communication pathways
Kim, Yelim Erin (Virginia Tech, 2026-06-01)
Tourism and hospitality operate in environments characterized by uncertainty, intangibility, and information asymmetry. In such settings, outside stakeholders often rely on visible cues to interpret future firm prospects and revise their expectations about value. This dissertation examines how such signals become economically meaningful through their connection to market valuation. Drawing on signaling theory and related perspectives on brand equity, strategic flexibility, and leadership, it investigates whether signals matter for valuation, when they matter more, and how firms can strengthen their valuation impact. By focusing on market valuation, this dissertation highlights how destinations and firms are evaluated not only through realized operating outcomes but also through the way external audiences interpret new information and translate it into expectations about future performance. Chapter 1 introduces the dissertation by outlining the role of signals in tourism and hospitality and presenting the broader theoretical and methodological framework. It explains why signals are especially important in industries where products and experiences cannot be fully assessed in advance and where destinations and firms operate under conditions of uncertainty. The chapter further develops the dissertation's central argument that signals may originate from outside the firm or from the firm itself, but their economic importance depends on whether they are noticed, interpreted favorably, and translated into value-relevant expectations. It also positions market valuation as the main outcome of interest and introduces the event study approach as a consistent methodological lens across the three studies. Chapter 2 examines whether destination-level popular culture functions as an external signal that enhances the market value of tourism-related firms through destination spillovers and the brand equity pathway. Drawing on customer-based brand equity and related perspectives on destination image and associations, this chapter argues that successful popular culture can strengthen destination visibility, enrich destination-related meanings, and increase the likelihood that the destination will be favorably considered by outside audiences. These effects may extend beyond tourism demand itself and influence investor expectations about firms operating within the destination. The findings suggest that popular culture can generate positive valuation effects by increasing destination visibility, strengthening destination-related associations, and shaping expectations about future tourism demand and firm prospects. Chapter 3 investigates when such external signals matter more by examining whether the valuation benefits of the same cultural signal vary across firms depending on organizational strategy. Although firms may be exposed to the same destination-level signal, they are not equally positioned to capture their benefits. This chapter focuses on international hotel firms and argues that operating arrangements differ in their ability to convert external signals into stronger market responses. Drawing on strategic flexibility and related perspectives on value capture, the chapter proposes that firms with more flexible structures are better able to respond to opportunities created by favorable destination-level attention. The findings show that franchised firms exhibit stronger abnormal returns than managed firms, suggesting that organizational strategy conditions the valuation effect of external cultural signals. Chapter 4 examines firm-initiated sustainability signals by focusing on environmental certification announcements in publicly traded U.S. hotel firms. While the earlier chapters focus on external signals, this chapter turns to signals generated by the firm itself and asks how such signals can become more effective in the eyes of outside stakeholders. The chapter argues that sustainability initiatives do not produce uniform valuation effects because market reactions depend on whether the signal is perceived as authentic and credible. Building on leadership-related perspectives, it proposes that the alignment between CEO political ideology and the firm's sustainability message strengthens the persuasiveness of the signal by reducing doubts about opportunism or greenwashing. The findings show that environmental certifications generate positive market reactions and that these effects are stronger when the CEO's political ideology is more closely aligned with the firm's sustainability message, highlighting the importance of perceived authenticity and credibility in shaping investor responses. The final chapter summarizes the main contributions of the dissertation and discusses implications, limitations, and directions for future research. Overall, this dissertation shows that signals do not affect valuation automatically. Their effects depend on whether they are recognized by outside audiences, whether firms are strategically positioned to benefit from them, and whether the signals appear credible and consistent with the broader context in which they are interpreted.
Generative AI for Hardening Security: From Training Data Augmentation, Language Alignment, to Code Vulnerability Demonstration
Kanchi, Shravya (Virginia Tech, 2026-06-01)
Machine learning-based cybersecurity systems face fundamental data limitations that constrain their effectiveness across diverse threat environments. Training data for security classifiers is inherently scarce, imbalanced, and subject to distributional shift, degrading generalization under evolving threat landscapes. Foundation model customization pipelines are vulnerable to data poisoning that induces harmful behavioral outputs. Vulnerability detection tools identify risks in software dependencies but provide no mechanism for determining whether reported vulnerabilities are concretely exploitable within a specific application context. This dissertation presents three contributions that leverage generative AI to address these challenges across training data augmentation, language alignment, and code vulnerability demonstration. First, we characterize the data challenges most commonly degrading ML-based security classifier performance through a measurement study of 35 papers published in top security venues. Building on this, we propose Nimai, a controlled synthetic data generation framework based on a Vector-Quantized Variational Autoencoder, designed to augment underrepresented regions of security feature spaces. Evaluated across seven security classification tasks, Nimai improves classifier performance in four of seven tasks, achieving up to 32 percent accuracy improvement under severe data constraints, and recovers up to 60.4 percent of performance degraded by concept drift. Second, we investigate the security of foundation model customization pipelines under toxicity injection attacks. We propose a healing data approach that employs a safety-aligned large language model to generate contextually relevant, prosocial replacements for identified toxic samples, actively instilling desirable behavioral alignment rather than discarding harmful content. Fine-tuning on the healed dataset achieves near-zero Response Toxicity Rates across three model architectures while preserving conversational quality, degrading gracefully under imperfect toxicity classifiers. Third, we address the challenge of bridging vulnerability detection and real-world exploitability assessment in software supply chains. We propose PoVSmith, an agent-based framework for automated proof-of-vulnerability test generation in Java applications. PoVSmith identifies application-level entry points to vulnerable library APIs through agent-based call path analysis, then iteratively generates, executes, and refines JUnit test cases guided by exemplar tests and execution feedback. Evaluated on 33 Java application-library pairs, PoVSmith reveals 158 unique application-level entry points to vulnerable library APIs and correctly identifies 152 of them with their call paths, achieving 96 percent precision. Using this call information, PoVSmith generates 152 PoV tests, of which 84, or 55 percent, demonstrate feasible ways of exploiting library vulnerabilities through application code. Collectively, these contributions demonstrate that generative AI can systematically strengthen cybersecurity pipelines by augmenting scarce training data to improve classifier generalization, curating conversational fine-tuning data to enforce safe model behavior, and producing executable evidence of real-world vulnerability exploitability, establishing generative AI as a unifying instrument for hardening security systems across data, model, and system layers.
AI-Driven Affective Captioning for Equitable STEM Access Among Deaf and Hard-of-Hearing Students
Ubur, Sunday David (Virginia Tech, 2026-06-01)
This dissertation investigates how Artificial Intelligebce (AI)- and Augmented Reality (AR)-supported captioning can improve communication access for Deaf and Hard-of-Hearing (DHH) learners in STEM contexts. Traditional real-time captions provide essential access to spoken language, but they often omit nonverbal and contextual information such as speaker identity, tone, emphasis, affect, and conversational intent. Across a preliminary design study and three empirical studies, this work examines how caption augmentations can preserve these missing layers of meaning while maintaining readability, timeliness, trust, and user control. The preliminary study compared traditional captions with emotion-augmented caption designs and showed that affective and visual cues can support comprehension when they are lightweight and text-centered, but may increase workload when they compete with the main transcript or visual scene. Study 1, a qualitative study with DHH participants, found that users valued emotion-aware captions when they clarified tone, emphasis, or speaker intent, but only when cues were timely, legible, optional, and subordinate to the transcript. Study 2 evaluated culturally adaptive emotive captioning in AR by comparing two cue formats: compact symbolic cues, implemented as emoji/icon indicators, and explicit textual affect labels, implemented as inline text-tags, across high- and low-context cultural cohorts. Compact symbolic cues produced a robust cross-cultural preference, while qualitative findings showed that participants valued the cues differently: some emphasized speed and reduced distraction, while others emphasized easier access to speaker emotion. Study 3 evaluated Speaker-Aware Affective Captioning, a multi-speaker captioning interface that combined speaker-attributed captions, confidence-gated affect tags, and an on-demand AI Describe feature. The study showed that speaker attribution was the most consistently valued support, while AI Describe helped users recover from missed or unclear information. Affect tags showed promise, but their usefulness depended on timing, persistence, interpretability, and trust. Across these studies, findings show that accessible captioning should not simply add more expressive information. Instead, next-generation captioning systems should reduce users' inferential burden through layered support: preserving the transcript first, identifying speakers, supporting recovery from missed information, and adding affective interpretation only when it is accurate, low-burden, and user-controllable. This dissertation contributes empirical evidence and design guidelines for trustworthy, culturally sensitive, and readable affective captioning systems for inclusive STEM learning.
Linking Semiochemical Ecology and Spatial Dynamics to Improve Integrated Pest Management of Striped Cucumber Beetle, Acalymma vittatum
Nunez, Demian Antonio (Virginia Tech, 2026-05-29)
The striped cucumber beetle, Acalymma vittatum (Fabr.), is a major early-season pest of cucurbits in eastern North America because overwintering adults injure seedlings while they are young and highly vulnerable to feeding damage or infection by Erwinia tracheiphila, the causal agent of bacterial wilt. This early vulnerability makes it important to reduce beetle activity before aggregations develop on young crop plants. This dissertation examined striped cucumber beetle semiochemical attraction, spatial capture patterns, baited-trap operating range, and early-season perimeter trap deployment to inform future selective semiochemical-based management strategies. The first study compared semiochemical lures for striped cucumber beetle attraction and non-target bee captures. Treatments containing vittatalactone, the A. vittatum aggregation pheromone, consistently produced the highest striped cucumber beetle captures while maintaining low bee bycatch. Pairing vittatalactone with indole as a companion lure may further improve A. vittatum captures without increasing bycatch. The second study examined how vittatalactone-baited trap captures were distributed across farm landscapes. A. vittatum captures were generally higher nearer tree lines, suggesting that placing traps along wooded field edges may improve early-season interception. The third study used fluorescent-marked beetles released at different distances from a vittatalactone + indole-baited sticky trap to estimate its operational range. Predicted recapture among dispersing beetles fell to 1% at approximately 12 m, suggesting that trap-out will likely require relatively dense trap arrays to maintain effective coverage. The final field study tested whether early-season perimeter trapping with vittatalactone + indole-baited sticky traps could reduce beetle activity within production cucurbit fields. This interval was chosen to intercept overwintered adults before crop-associated cues could compete strongly with traps and to reduce early aggregation on vulnerable seedlings. Trap-out sites had lower striped cucumber beetle activity during crop establishment and early crop development in 2024, although this effect was not repeated in 2025. Together, these findings show that semiochemical trap-out can reduce striped cucumber beetle pressure under some field conditions and, with refinement, could support a viable early-season management strategy.
Evaluating the Effect of Mating Disruption and Parasitoids on Management of Diamondback Moth, Plutella xylostella (L.), in Commercial Brassica Systems
Tomlinson, Taylore Ashley (Virginia Tech, 2026-05-29)
The diamondback moth, Plutella xylostella (L.) (Lepidoptera: Plutellidae) has become increasingly difficult to manage with insecticides alone, since it has developed insecticide resistance in recent decades. In the United States, brassica producers are running out of management options for controlling P. xylostella, resulting in a need for implementation of integrated pest management (IPM) tactics, such as mating disruption, biological control, and rotation of chemical controls. From 2021-2025, trials in brassica systems examining the effects of pheromone use and mating disruption on P. xylostella were conducted in four different states. Hand-applied mating disruption dispensers were found to reduce P. xylostella moth captures in treated plots during the growing season. It is also important to note the biological control agents have played an active role in targeting P. xylostella. As a result of in-field parasitism surveys conducted across Virginia from 2022-2025, Diadegma insulare (Hymenoptera: Ichneumonidae) was found parasitizing P. xylostella at moderate rates. Decreasing the number of insecticide spray applications to control P. xylostella is important to conserve natural enemies and to reduce the risk of insecticide resistance development. In 2025, a case study utilizing threshold-based scouting for lepidopteran pests paired with mating disruption in commercial cabbage fields showed a reduction of at least two insecticide spray applications spanning over 100 acres in southern Virginia. These trials provide promising results for P. xylostella management in the eastern United States.
Universal Design for Learning Strategies to Improve Student Engagement, Confidence, and Participation in an Undergraduate Biochemistry Laboratory Course
Drolet, Erin Taylor (Virginia Tech, 2026-05-29)
Biochemistry laboratory courses present unique challenges to students, especially those with disabilities. This dissertation investigates how inquiry-based learning (IBL) and Universal Design for Learning (UDL) can be systematically integrated into an undergraduate biochemistry laboratory course to address these challenges. This work evaluates curricular, instructional, and environmental modifications in a longitudinal study. Aim 1 demonstrates that a module-based laboratory course can be reorganized into a inquiry-based format while preserving core technical competencies, resulting in high levels of student engagement and perceived gains in durable skills. Aim 2 applies UDL principles of Representation, informed by cognitive load theory and text-signaling, to redesign laboratory manuals, leading to reduced student stress and improved navigation of course materials while maintaining learning outcomes. Aim 3 examines the physical accessibility of laboratory environments through student feedback, highlighting how they influence participation and identifying both improvements and persistent barriers across different instructional spaces. Collectively, these findings contribute to the literature by bridging research on undergraduate laboratory courses and UDL in biochemistry, demonstrating that scalable, low-cost interventions can improve engagement and accessibility in laboratory settings. This work provides an evidence-based framework for integrating inquiry, accessibility, and skill development in biochemistry laboratory education and underscores the importance of aligning instructional design with both cognitive and physical dimensions of student experience.
From Step Tests to Soft Sensors: Model-Informed Controller Tuning and Hybrid Feedforward–Feedback Ammonia-Based Aeration Control to Improve Full-Scale Water Resource Recovery Facility Performance
Gagnon, Alexandria Augusta (Virginia Tech, 2026-05-29)
Aeration is essential for nitrification and biological nitrogen removal in water resource recovery facilities (WRRFs), but it is also one of the largest operating energy demands. As utilities face increasingly stringent effluent nitrogen limits and pursue greater energy efficiency, aeration control strategies that stabilize effluent ammonia while minimizing unnecessary oxygen supply have become an operational priority. Ammonia-based aeration control (ABAC) is an effective approach because it links dissolved oxygen targets to real-time ammonia measurements. However, in many full-scale facilities, ABAC and other feedback loops are constrained by controller tuning practices based on trial-and-error or ad hoc rules often producing sluggish or oscillatory behavior, and reduced operator confidence. This work develops and validates practical methods to systematically tune proportional–integral (PI) controllers and improve ABAC performance under realistic WRRF dynamics. Because PI control remains the dominant structure in treatment plant automation and is well suited to first-order-plus-deadtime (FOPDT) processes, this research focuses on methods that characterize loop dynamics with minimal plant disruption, generate repeatable tuning parameters, and identify when feedback-only ABAC must be augmented with predictive action. Open-loop step-response testing was first applied to representative WRRF control loops to develop FOPDT models relating manipulated and controlled variables. These models were then used with lambda tuning to compute PI parameters that balance stability and responsiveness for both fast loops, such as airflow and header pressure control, and slower nutrient-related process loops. Because step-response testing is often impractical for ABAC under variable full-scale conditions, a reduced-order model-based tuning method was then developed. A hydraulics-based reduced-order model with simplified activated sludge relationships was used to describe how oxygen availability influences nitrification capacity. The non-linearity associated with Monod saturation kinetics was explicitly integrated into the controller structure so that feedback action operated on a more linearized response surface. Monod saturation nonlinearity was incorporated into the controller structure so feedback acted on a more linearized response surface. In model-based validation, the kinetic-informed controller achieved a mean absolute error (MAE) of 0.09 mg N/L relative to the effluent ammonia setpoint. When implemented at full scale and tuned using the proposed method, the controller achieved stable operation and a 0.16 mg N/L MAE. This work also addressed facilities with plug-flow hydraulics and pronounced diurnal loading, where feedback-only ABAC can become deadtime-dominant and respond only after a disturbance has propagated through the aeration system. Frequency-response screening was used to evaluate controllability limits imposed by transport delay and to identify conditions where feedback tuning alone is insufficient. A hybrid feedforward-feedback ABAC (FFABAC) architecture was then implemented in which model-derived soft sensors forecast influent ammonia loading and nitrification capacity for proactive feedforward action, while PI feedback corrects model error and maintains long-term stability. FFABAC achieved an overall MAE of 0.22 mg N/L and improved to 0.16 mg N/L under unconstrained conditions. In a controlled comparison, FFABAC reduced ammonia MAE from 0.31 ± 0.28 mg N/L under feedback-only ABAC to 0.11 ± 0.10 mg N/L. Collectively, this work provides a practical, scalable toolkit for improving aeration-related control performance in WRRFs. Step-response testing with lambda tuning offers a repeatable method for tuning common PI loops, reduced-order model-based tuning provides a feasible pathway for ABAC where traditional step tests are impractical, and frequency-response screening offers a framework for when feedback should be augmented with predictive feedforward control. These methods are designed for implementation within standard plant automation platforms, enabling systematic tuning, measurable performance improvement, and reduced operational risk.
CEO Power and Lead Independent Directors
Ding, Weiming (Virginia Tech, 2026-05-29)
Although prior research documents a positive association between CEO duality and the presence of a lead independent director (LID), I show that this finding may be incomplete. Conditional on duality, concentrated CEO power, measured using a multidimensional index of pay slice, ownership, and tenure, is negatively associated with LID presence and with LID presence on key monitoring committees. These findings are consistent with managerial power theory: CEOs leverage their power to resist independent board leadership, and this resistance is consequential. LIDs who serve on the nomination committee are meaningfully associated with increased sensitivity of CEO forced turnover to poor performance, and financially expert LIDs are associated with strong oversight outcomes more broadly. I further show that once an LID is appointed, CEO power is associated with a lower likelihood of LID discontinuance, suggesting that the role becomes entrenched over time. Collectively, my findings reveal that CEO power influences not only LID presence but also the conditions under which the role is effective. These results have direct implications for investors and practitioners seeking to strengthen independent board leadership.
Polymer Phase Behavior and Morphology Control: From Blend Compatibilization to Aerogel Formation
Trindade Coutinho, Isabela (Virginia Tech, 2026-05-29)
This dissertation covers polymer blend compatibilization and improvements in the mechanical properties of semicrystalline polymer aerogels. Both blends and aerogel properties are governed by phase separation processes. While in polymer blends, domain size and interfacial adhesion related to the phase separation can be detrimental for the properties, in aerogels obtained through thermally induced phase separation (TIPS), the phase separation is responsible for the network formation. Therefore, in both areas, controlling the phase behavior is necessary to control the properties of the final blend and aerogels. In the first half of this dissertation, the compatibilization of blends of polysaccharides with polyesters is investigated. Polymer blends are the physical mixture of at least two polymers, which are designed to achieve improved properties compared to the pure polymers. While polysaccharides are sustainable polymers, their use as plastics is limited due to frequent low toughness, poor melt processability, high water sensitivity, and high production costs. Therefore, making blends of polysaccharides is an alternative to mitigate this shortcoming and expand the use of sustainable polymers. Due to a small entropy of mixing, polymer blends tend two phase separate. The phase-separated morphology is characterized by sharp interfaces with low adhesion, which leads to poor properties. To mitigate the consequences of phase separation, compatibilization is achieved by adding compatibilizers that favorably interact with both components in the blend. Chapter 1 discuss the fundamentals of polymer blends and polymer blends compatibilization. Chapter 3 and Chapter 4 investigate the use of a block polymer and a graft polymer, respectively, in the compatibilization of polysaccharide/polyester blends to advance the knowledge about blend compatibilization. In both chapters, phase contrast optical microscopy (PCOM) and small-angle laser-light scattering (SALLS) were used to track changes in the phase-separated morphology as the compatibilizers were added. In Chapter 3, ethyl cellulose (ECel)/poly(ethylene terephthalate) (PET) 70/30 blends were compatibilized with a block polymer of ethyl cellulose (ECel) and poly(benzyl glutamate), named ECel-block-poly(BG). Different amounts of the block polymer, 5, 10, 20, and 30 wt.%, were tested as compatibilizers. The uncompatibilized blend presented a highly phase-separated morphology composed of large and small domains, characteristic of late stages of spinodal decomposition. As the compatibilizer content increased, the size of the large domains decreased until a bi-continuous spinodal texture was obtained with 30 wt% of compatibilizer. A decrease in average domain size from 15 ± 4 μm in the uncompatibilized blend to 2 ± 1 μm when using 30 wt% of the compatibilizer was observed. These changes in domain size highlight the ability of the compatibilizer to steric stabilize these blends, preventing coarsening of the phase-separated morphology. Chapter 4 investigated the impact of amylose acetate-graft-poly(D,L-lactic acid) (AmAc g-PDLLA) graft density and graft length on the compatibilization of starch acetate/PDLLA 70/30 blends. Graft polymer contents of 5, 10, and 20 wt% with varying graft density and graft length were investigated. The results showed that in order for the compatibilizer to reduce the interdomain distance of the blend, it has to entangle with the polymers of the blend. Furthermore, the ability of the compatibilizer to entangle was related to the chain entanglement molecular weight (Me) of the polymers in the blend. The series of graft polymers with the same graft length (29.4 kg/mol) but different graft densities (between 0.5 and 19 %) showed that the graft density has to be low enough so that the segments between grafts are at least the Me of the starch acetate, allowing the graft polymer to entangle with the starch acetate and promote compatibilization. In parallel, the series of graft polymers with the same graft density (1 %) but different graft lengths (between 7.9 and 29.4 kg/mol) showed that the grafts have to have a molecular weight above the Me for PLA, allowing the grafts to entangle with PDLLA and promote compatibilization. The second half of this dissertation investigates semicrystalline polymer aerogels obtained through TIPS. On TIPS, the polymer is dissolved at a high temperature, and upon cooling, the gel network is formed through a phase separation process. Chapter 2 discusses the fundamentals of the TIPS process and polymer aerogels. For the systems investigated here, the phase separation happens through solid-liquid phase separation, where the polymer crystallizes from the solution to yield the aerogel network. On TIPS, there are many parameters that can be tuned to control the morphology of the aerogel, including the initial polymer content, the solvent, the dissolution temperature, the gelation temperature, and the presence of additives. In Chapter 5, the impact of the gelation temperature on the properties of poly(ether ether ketone) (PEEK) aerogels was studied. It was observed that by increasing the gelation temperature the aerogel network connectivity was enhanced. As the mechanical properties of aerogels depend on the network connectivity, this increase in connectivity resulted in up to a 111.5 % improvement in the compressive modulus of the aerogels, while crystallinity, density, and porosity remained unchanged. Chapter 6 explored another approach to improve the mechanical properties of aerogels while maintaining their porosity. Specifically, the incorporation of sodium montmorillonite, Cloisite 10A, and Cloisite 25A nanoclays to polyphenylene sulfide (PPS) aerogels was investigated. The addition of 1 wt% of any type nanoclay did not impact the morphology of the PPS aerogel, but adding 5 wt% of Cloisite 10A resulted in a less connected morphology and therefore worse mechanical properties. An increase of 30 % in the compressive modulus of the aerogel was observed when 1 wt% of the montmorillonite was added to 15 wt% PPS aerogels. While some intercalation was observed for montmorillonite, we believe that the compressive modulus was not further enhanced because the nanoclay was not fully exfoliated in the aerogel matrix. Lower improvements in compressive modulus were observed by the addition of Cloisite 10A and Cloisite 25A, which were related to the degradation of the organic modifier. The degradation of the organic modifier can lead to a worse distribution of the nanoclays in the polymeric matrix, which is detrimental to mechanical properties. Some dependence on the nature of the organic modifier was also observed, highlighted by the better performance of Cloisite 10A compared to Cloisite 25A. Finally, the degradation temperature of the aerogels was increased by the addition of the nanoclays.
Flood pulse effects on multispecies catch in the Amazon Basin
Borba, Gabriel Costa (Virginia Tech, 2026-05-29)
Tropical river-floodplain fisheries feed tens of millions of people and depend on the flood pulse, the seasonal rise and fall of water that drives fish recruitment, growth, and catch. Yet we cannot predict how catch will respond to changing flood regimes. Evidence is fragmented; most models emphasize high-water dynamics while ignoring low-water dynamics, and species-specific responses are rarely linked to the life-history traits that shape them. These gaps matter, as dams, deforestation, and climate change reshape flood pulses worldwide. This dissertation builds a mechanistic, trait-based, and climate-forward understanding of the flood pulse–fish catch relationship in the Brazilian Amazon. I synthesized evidence from 27 studies across tropical basins. I then paired 20 years of daily fish landings (1991–2011) from 11 Amazon ports with daily water levels to identify which flood pulse features drive catch, tested whether life-history traits moderate taxon-level responses across 14 dominant taxa, and projected catch to 2100 under climate scenarios. A composite hydrological signature capturing prolonged high water, elevated minimums, and gradual rises, lagged two years, predicted catch better than any single metric, explaining 78% of the deviance. Size at maturity most strongly moderates species responses: early-maturing taxa lose up to 22% of catch under high flood conditions, while later-maturing taxa gain up to 11%. Under mid-century climate projections, basin-wide catch appears nearly stable, but this aggregate conceals near-universal declines across 8 of 9 regions and all 14 taxa, concentrated in the drying tributaries. The flood pulse–fish catch relationship is mechanistic, trait-filtered, and forecastable, but basin-wide metrics mask the regional and taxonomic losses where climate impacts are felt. Anticipating impacts in river-floodplain fisheries requires standard hydrological metrics, tributary-scale projections, trait-specific models, and management strategies that match each species' flood sensitivity.
Kernel-Based Metamodeling for Heteroscedastic Simulation: Theory and Distribution-Free Uncertainty Quantification
Zhao, Jin (Virginia Tech, 2026-05-28)
Simulation metamodels approximate expensive stochastic simulators for downstream analysis and decision-making. They can become unreliable under strong heteroscedasticity, large or high-dimensional datasets, and tight simulation budgets for uncertainty quantification. This dissertation develops scalable prediction, convergence theory, and distribution-free uncertainty quantification. Nested heteroscedastic Gaussian process (NHGP) provides a scalable divide-and-conquer scheme: each disjoint subset fits a sub-stochastic-kriging model, and the fits are fused into one metamodel. Theoretical results show that the NHGP rule is the best linear unbiased predictor among such aggregations, prove consistency, and demonstrate empirical performance comparable to a benchmark method. A second thread studies stochastic-kriging prediction error under kernel misspecification and growing dimension: it characterizes near-optimal regimes, explains how noise-variance scaling governs worst-case performance, and proves high-probability pointwise bounds whose dimension dependence appears only through a log n factor. The third methodological piece is heteroscedastic weighted kernel ridge regression, which uses precision-based weights for input-dependent noise. Using the equivalence between Gaussian-process prediction and kernel ridge regression, we derive sharp mean squared error rates across kernel spectrum decay regimes, identify rate-optimal regularization, and validate the theory empirically. Finally, we combine conditional kernel mean embeddings with distributional conformal prediction to estimate conditional cumulative distribution functions and build calibrated prediction sets with finite-sample coverage, together with an adaptive two-stage budget allocation between estimation and calibration.
Unraveling Gene Regulation in Sea Urchins and Resurrection Plants through Integrative Network-Based Approaches
Zhang, Jingyi (Virginia Tech, 2026-05-28)
Gene regulatory networks (GRNs) coordinate gene expression by defining the complex regulatory dynamics between transcription factors and their target genes. While high-throughput sequencing technologies provide dense temporal expression data, computationally reconstructing accurate network topologies remains an optimization challenge due to inherent noise, high dimensionality, and the dynamic nature of biological systems. This dissertation integrates advanced computational frameworks to infer, analyze, and align dynamic GRNs across diverse biological models. Specifically, this work contributes to three computational aspects of systems biology: (1) the development and validation of a predictive data processing pipeline using Strongylocentrotus purpuratus (sea urchin) temporal expression data, (2) the application of ensemble inference and graph representation learning to decode the regulatory topology of vegetative desiccation tolerance in Selaginella sellowii, and (3) the formulation of a relaxed topological alignment algorithm to discover conserved regulatory modules across divergent plant species. Results demonstrate that robust machine learning architectures can successfully predict complex temporal GRNs. The initial pipeline evaluation achieved high predictive sensitivity in reconstructing the sea urchin endomesoderm and ectoderm networks, relying solely on temporal expression data without replicates. Furthermore, to analyze the dynamic stress response in S. sellowii, an advanced architecture integrating a two-level stacking ensemble classifier with graph attention networks (GAT) and protein language model embeddings was developed. This approach revealed a unique survival strategy, where the regulatory network undergoes a structural contraction into a minimalist stable state under extreme dehydration. This transition is coordinated by a core set of persistent regulatory hubs that subsequently undergo context-specific functional rewiring to orchestrate rapid recovery. Finally, addressing the computationally intractable network alignment problem, a novel greedy algorithm bounded by a dynamic $epsilon$-stopping condition was designed to navigate complex many-to-many orthology mappings. By relaxing the constraints of strict topological isomorphism, this approach successfully extracted a heavily connected, conserved regulatory module across Arabidopsis thaliana, Zea mays, and Sorghum bicolor, linking conserved upstream regulators to complex, species-specific downstream execution pathways. Overall, this dissertation establishes robust and scalable computational methodologies for extracting hierarchical regulatory architectures from noisy transcriptomic data sets. These frameworks provide a rigorous, data-driven blueprint for identifying core regulatory logic across complex biological systems, facilitating the translation of conserved network architectures among species.
Making Sense of Regenerative Tourism: From Inspiration to Institutionalization through Social Movements, Framing and Co-Creation
Hajarrahmah, Dini (Virginia Tech, 2026-05-28)
This three-article dissertation examines strategies to advance regenerative tourism across key stakeholders in multiple ways. Drawing on social movement theory, Study 1 examines how 57 tourism social entrepreneurs across six continents navigate challenges, opportunities, and strategies to advance regenerative tourism through in-depth interviews. This work identifies systems thinking and feedback loops as critical ingredients for regenerative tourism. Study 2 investigates how framing strategies influence visitors' perceptions and intentions to participate in regenerative tourism through a sequential mixed-methods design (integrating in-depth interviews with DMO leaders and regenerative practitioners and an online experiment in the U.S., the U.K., and New Zealand). Findings show that prognostic and motivational framing significantly shape tourists' attitudes and intentions to participate in regenerative tourism, especially among individuals with strong pro-social identities. Study 3 explores how regenerative agritourism is designed, diffused, and adopted through co-creation. This study employs a multi-method qualitative design, integrating field observations of regenerative farm tours and in-depth interviews with key stakeholders (e.g., regenerative farmers, visitors, and supporting organizations). Findings indicate that innovation is not linear but shaped by iterative, multi-level feedback loops. Visitors are not merely passive adopters but active contributors through hands-on stewardship experiences. Overall, this dissertation illuminates the regenerative tourism process, making sense of it from the inspiration stage through the institutionalization of broader regenerative practices.
Embracing Chaos: Exploring Conformational Landscapes of Intrinsically Disordered Proteins
Gil Pineda, Laura Isabel (Virginia Tech, 2026-05-28)
Intrinsically disordered proteins (IDPs) play central roles in cellular signaling and regulation, yet their dynamic and heterogeneous conformational landscapes present significant challenges for experimental and computational characterization. These challenges are further amplified by processes such as post-translational modification and aggregation, which can reshape IDP conformational ensembles and are often linked to disease. Molecular dynamics simulations offer a powerful framework for studying IDPs at atomic resolution; however, their accuracy depends critically on the quality of the underlying force fields and the ability to efficiently sample complex free-energy landscapes. This dissertation addresses these challenges through the development and application of computational methods to study IDPs across multiple levels of complexity. First, the Drude polarizable force field is extended through the parametrization and validation of phosphorylated serine, threonine, and tyrosine residues, enabling more accurate representation of electrostatic effects associated with phosphorylation. Next, enhanced sampling strategies are evaluated using the β-catenin17−48 peptide as a model system to determine how different methods influence the exploration of phosphorylation- dependent conformational landscapes and the identification of intermediate states. Using one of the enhanced sampling strategies, aggregation of the amyloid-β fragment Aβ16−22 is investigated revealing distinct aggregation pathways, heterogeneous oligomeric intermediates, and the role of hydrophobic interactions in stabilizing early assemblies. Finally, the complex IDP α-synuclein is examined, focusing on how phosphorylation alters the electrostatic and structural properties of fibrillar assemblies. Together, this work establishes a computational framework for studying IDPs that integrates improved force field accuracy with enhanced sampling methodologies. These approaches provide insight into how phosphorylation and aggregation reshape conformational landscapes and offer a foundation for investigating the molecular mechanisms that underlie IDP function and dysfunction in disease.
Investigating Immersive Collaboration Across Temporal States
Giovannelli, Alexander (Virginia Tech, 2026-05-28)
As technological advancements continue to enable globally distributed work, teams are increasingly comprised of experts who are geographically separated. While telepresence solutions such as Zoom and Microsoft Teams enable remote collaboration among team members, these applications do not afford a common space for shared activity comparable to a physical workplace. Collaborative Virtual Environments (CVEs) represent a next generation of telepresence solutions that enable geographically distributed teams to collaborate within a shared digital workspace analogous to an in-person office. Team members are embodied in these CVEs through Head-Worn Displays (HWDs), which capture user movements and behaviors using various sensors and replicate them onto their virtual representations: avatars. By leveraging CVEs and their capacity to represent and communicate user actions, cooperative work can occur in ways akin to co-located collaboration while removing geographical constraints. However, research must examine how to support this telepresence modality across temporal contexts for it to meaningfully support collaboration. This dissertation examines four CVE research efforts to determine best practices for supporting collaboration across time. Specifically, it contributes by: (1) exploring the use of synthetic visuals to augment nonverbal expressions during synchronous collaboration; (2) investigating interactive guided tours with natural annotations (i.e., avatar recordings) to facilitate change awareness and information recall during asynchronous collaboration; (3) identifying the potentials and limitations of each temporal state in XR collaboration, including the introduction and examination of bichronous collaboration as a previously underexplored temporal state; and (4) evaluating the use of emerging technologies, specifically conversational agents, to support question–answer exchanges when live collaborators are unavailable, thereby aiding interpretation of prior contributions within a CVE.
Physiological and Molecular Approaches Revealing Mechanisms of Fruit Abscission and Pre-Harvest Drop Management in Apple (Malus × domestica Borkh.)
Tipu, Mohammad Monirul Hasan (Virginia Tech, 2026-05-28)
Pre-harvest fruit drop (PFD) is a major constraint on yield and profitability in commercial apple (Malus × domestica Borkh.) production, with susceptible cultivars like 'Honeycrisp' often experiencing significant losses. While current management using ethylene inhibitors such as aminoethoxyvinylglycine (AVG) effectively reduces PFD, it often delays ripening, suppresses anthocyanin biosynthesis, and degrades overall market quality. In this study, the efficacy of various plant growth regulators (PGRs), including ACC (1-aminocyclopropane-1-carboxylic acid), ethephon, and AVG, was first evaluated over two consecutive years (2023–2024). The integrated application of ACC+AVG was found to be the most effective strategy, reducing PFD by 27.05–46.30% while simultaneously intensifying red coloration by upregulating anthocyanin biosynthetic genes such as MdCHS, MdCHI, and MdDFR. Because this treatment showed no significant shift in internal ethylene concentration (IEC) compared to the untreated control, further investigation was conducted to determine whether fruit abscission and ripening are biologically coupled or regulated independently. Subsequently, IEC and key maturity indices—including firmness, soluble solids content (Brix), and starch index—were compared between dropped and retained fruits after gentle agitation at the same phenological stage. These indices were found to be statistically indistinguishable, indicating that fruit abscission can occur independently of ripening progression. This confirmed the functional decoupling between the two processes and prompted an investigation into the specific tissue-level site where this regulation occurs. Transcriptomic profiling of the pedicel abscission zone (AZ) and the fruit cortex was then performed across two critical phenological windows: one week before anticipated commercial harvest and one-week post-harvest. A profound tissue-specific divergence was identified, with 848 differentially expressed genes (DEGs) in the pedicel compared to only 21 in the cortex, establishing the pedicel as the primary site of abscission regulation. To narrow the focus further, pedicel-specific gene modules were characterized and identified as responsible for coordinated shifts in auxin signaling and oxidative stress during abscission. Based on these molecular findings, a translational field study was conducted to test if targeting these specific pathways could control PFD without affecting ethylene dynamics. Field applications of synthetic auxin (NAA) and the antioxidant melatonin were found to significantly reduce PFD (23.30% and 26.08%, respectively) relative to the control. Notably, melatonin induced expression of the peroxisomal protease MdLON2 in the pedicel, suggesting that oxidative homeostasis is a critical regulator of the abscission process. Together, these results establish a tissue-specific framework for fruit abscission that is functionally independent of ripening. By demonstrating that the pedicel acts as the primary "control center" for dropping, this work provides a translational basis for targeted interventions that extend the harvest window while maintaining the quality of 'Honeycrisp' apples.
Learning to Collaborate: Toward Robust, Adaptive Policies for Human–Robot Teams
Sagheb, Shahabedin (Virginia Tech, 2026-05-27)
Robotics, automation, and the use of Machine Learning (ML) algorithms have been steadily making progress. They have been adopted in various sectors, including manufacturing, education, healthcare, and transportation. Although intelligent algorithms behind text-based, image-based, and commerce platforms are very prominent and often cited as examples of progress, there exists a gap in applying these algorithms to robotics applications with humans in the loop. In addition to human acceptance, robotic systems require safety-critical interfaces with humans (e.g., self-driving technology and robotic-assisted living). There is also a need for robot-specific datasets to train these algorithms. Providing efficient ways to train algorithms and building intuitive and safe interfaces with humans can lead to increased adoption and trust between end-users and robotic systems. The paradigm of Human--Robot Collaboration (collaborative autonomy) has been one of the most promising approaches to gathering data from human users and incrementally building trust between the human and the machine. However, humans are not static agents. Algorithms working with humans must consider the dynamic nature of their interactions with human users. This creates an exciting and challenging opportunity to develop algorithms that learn from humans and adapt to the requirements of an evolving task. In this dissertation, we investigate how robots can be trained efficiently and robustly given the dynamic nature of humans. Concretely, we explore three key objectives: (1) developing algorithms that learn efficiently from limited human demonstration datasets, (2) developing decision-making policies for long-term interaction, and (3) developing robot policies that communicate the learning to humans. This research leverages existing methods and builds on them to present novel approaches for learning from, communicating with, and adapting to human users. Our results are agnostic to the application domain (e.g., healthcare or driving) and to the type of robot (e.g., robot arm vs. autonomous car). Our main contributions are: (1) a learning algorithm for efficiently learning from human teachers, (2) a foundational optimization framework for influencing human partners over long-term interactions, and (3) a game-theoretic approach to communicating robot learning to human partners. We provide algorithms and experimental results from evaluations in simulated and real environments that demonstrate the effectiveness of our proposed approaches.

Browse

Recent Submissions