Biological and Health Sciences

Investigating the Design, Implementation, and Use of Large Language Model-Generated Draft Replies in Clinical Practice: An Ethnographic Study
Philip Barrison, Health Infrastructures and Learning Systems

As the healthcare industry attempts to address the growing administrative and documentation burden placed on healthcare providers, tools leveraging novel generative artificial intelligence (AI) technologies have been rapidly adopted to address this problem. These emerging technologies have the potential to reconfigure clinical work practices and how healthcare providers engage with their patients. In this dissertation, I investigate how one such intervention, generated draft replies (GDR), is implemented into clinical practice. Using ethnographic methods, including participant observation, interviews, and review of primary source documents, this research investigates how disparate social worlds collaborate when implementing novel AI technology into clinical practice. By critically examining how individuals from different backgrounds make meaning out of GDRs, collaborate to move their technology into practice, and how those decisions influence user adoption, this dissertation will significantly influence research, practice, and institutional policy decisions related to the expansion of novel AI technologies in patient care.

Cerebral Blood Flow, Perivascular Function, Clinical Presentation, Recovery, and Sleep Duration Following Sport-Related Concussion in National Collegiate Athletic Association Athletes
Adrian Boltz, Movement Science

Sport-related concussion (SRC) is a mild traumatic brain injury resultant of biomechanical forces to the head/neck, and causes physical-, cognitive-, behavior-, and sleep-impairment. Sleep duration is frequently impacted following SRC, however it has been understudied in the context of SRC sequalae. This dissertation will investigate the relationship between acute post-SRC sleep duration (within 24 to 48 hours) and acute (and longitudinal trajectory of) clinical presentation (balance, neurocognition, symptomology), neurophysiology (perivascular function, cerebral blood flow, and protein [GFAP, UCHL1, NFL, and Tau] concentration), and recovery time (days to initiation of return to play protocol and unrestricted return to play) in National Collegiate Athletic Association student-athletes. Accordingly, this dissertation aims to elucidate key mechanisms and patterns that can be leveraged to expand our understanding of the role sleep has in SRC pathophysiology and sequelae, support in the identification of individuals at risk for protracted recovery, consequently guiding tailored interventions in this vulnerable population.

Tryptophan Metabolism and Noncanonical AHR Signaling in Pulmonary Fibrosis
Hannah Carter, Microbiology and Immunology

Idiopathic pulmonary fibrosis is a progressive, lethal disease with no known cause and unclear immune involvement. We have found evidence that the aryl hydrocarbon receptor (AHR), despite initiating well-known anti-inflammatory signaling, is acting in a pro-inflammatory, profibrotic manner during fibrosis. We found that dendritic cells (DCs) express significant AHR, are hyperinflammatory in fibrotic lungs, and yet show no traditional AHR signaling. Treating mice with an AHR activator during the fibrotic timeframe of our model worsens fibrosis, while treating with an inhibitor is protective. We further found that both human and mouse fibroblasts from fibrotic lungs increase the production of a natural AHR activator via tryptophan metabolism and can lead to more inflammatory signaling from DCs. Inhibiting this excess production protected from fibrosis. Overall, we showed that in the fibrotic lung, fibroblast and DC cross-talk worsens inflammation and fibrosis through AHR.

On the Price of Land Back: (Re)Negotiating Relationships with Molokai Ranch
Kristofer Castro, Environment and Sustainability

Indigenous peoples have long advocated for the return of lands that have been taken from them. In environmental science, there is growing evidence that land back will produce important conservation and sustainability outcomes. Yet, few studies exist that cover how land should be feasibly priced for Indigenous peoples to acquire the land, especially in cases where Indigenous peoples aim to reestablish traditional conservation and subsistence practices and set up renewable energy and other sustainable systems. While pricing is often assumed to conflict with Indigenous values and worldviews, no study has yet examined how Indigenous peoples have managed the financial aspects of land back. This dissertation follows a case of land back on the island of Moloka’i to understand how the Indigenous community has strategically shaped pricing toward implementing a major land back project. This research contributes to growing literature on finance as a tool for land back.

A STAT5 Activator Promotes a CD8+ Memory T Cell Fate Independent of Cytokine
Monica Chanda, Cancer Biology

The ability of memory CD8+ T cells to self-renew and persist is critical to achieving efficacious cancer immunotherapies and long-term protection from pathogens. Yet, how cytokine-induced JAK-STAT signaling results in differentiation to a memory fate is unclear. Because T cell differentiating cytokines act through STAT5, but produce different fate outcomes, a strategy to activate STAT5 (aSTAT5) without the need for cytokines was developed. I performed RNA sequencing, qPCR, and metabolic analyses to compare aSTAT5 to memory and effector differentiating cytokines. I found aSTAT5 caused a memory-like CD8+ T cell fate. I also observed aSTAT5 increased memory T cell formation in mice that were infected with acute LCMV. Analysis of signaling pathways uncovered differences between aSTAT5 and cytokines, notably activation of the RAS-ERK pathway. My findings provide a therapeutic rationale for the use of direct activation of STAT5 to augment T cell therapies by promoting a memory T cell fate.

China’s Role in Shaping Energy Access in Africa
Congyi Dai, Environment and Sustainability

Energy poverty in sub-Saharan Africa severely hinders economic and social development. Providing modern energy services is not only important to the people of Africa, but also essential to the global energy transition. China, as a leader in energy technology manufacturing and critical mineral processing, is deeply involved in Africa’s energy and mining sectors. However, due to the opaque nature of Chinese financing, there is limited empirical evidence on its approach, intentions, and impacts. This dissertation examines China’s role in Africa’s energy access through: (1) the relationship between off-grid solar devices and development outcomes in rural Malawi; (2) the impacts of China’s energy projects on local communities’ energy access in Zambia; and (3) the links between China’s development aid and mine acquisition across sub-Saharan Africa. This work sheds light on how China is directly and indirectly influencing energy access at multiple scales and on sovereignty over resource governance and management.

Determining the Role of Therapy-Induced Senescence and Myeloid Plasticity in Multiple Myeloma
Angelo Guilatco, Cancer Biology

Multiple myeloma (MM) is an incurable plasma cell (PC) cancer in which virtually all patients relapse. Standard of care involves conditioning with high-dose melphalan (HDM), a cytotoxic therapy that can drive therapy-induced senescence (TIS). Using in vitro and in vivo models of HDM-MM, along with evaluation of MM patient biopsies, we have identified TIS features in MM. Interestingly, we have also identified a myeloid signature in HDM-MM cells and patient PCs. There is overlap between senescence and the integrated stress response (ISR), another stress response mechanism that signals through the CEBP transcription factors, which are known to regulate myeloid differentiation. Thus, we hypothesize that TIS activates myeloid plasticity in MM via the ISR/CEBP signaling to promote survival. We are further exploring the implications of MM myeloid plasticity, including altered interactions with the bone niche. Successful completion of our studies will justify targeting TIS and myeloid plasticity to eliminate growth-arrested MM.

Pan-cancer Analysis of Protein N-Glycosylation
Yi Hsiao, Bioinformatics

Protein glycosylation, where glycans are covalently linked to polypeptides, is a common post-translational modification (PTM) affecting various biological processes and increasingly linked to multiple cancer hallmarks. While large-scale analyses of PTMs like phosphorylation and acetylation have been advanced by efforts such as the Clinical Proteomic Tumor Analysis Consortium (CPTAC), glycosylation remains underexplored. Reanalyzing CPTAC’s protein phosphorylation data from over 1,400 tumor samples across nine cancer types enriched by immobilized metal affinity chromatography, we identified more than 96,000 co-enriched N-glycoforms. This comprehensive atlas of protein N-glycosylation for cancers enabled us to reveal novel pan-cancer biomarkers, therapeutic targets with clinical translation potential, and insights into mechanisms of dysregulated glycosylation. Our findings not only enhance understanding of cancer biology but also demonstrate the feasibility of glycosylation analysis using phosphorylation-enriched datasets previously collected, paving the way for similar studies to investigate other diseases.

Integration-Segregation Balance as a Neural Mechanism of Consciousness
Hyunwoo Jang, Neuroscience

This dissertation investigates how the brain’s network organization underpins consciousness, focusing on the dynamic interplay of integration and segregation. I introduce the integration-segregation difference (ISD), a novel neuroimaging metric that unifies network efficiency and clustering to quantify this balance. First, leveraging the Lifespan Human Connectome Project functional MRI dataset, I demonstrate ISD’s sensitivity to developmental and aging-related changes in brain network organization. Second, I apply ISD to characterize altered states of consciousness induced by anesthesia and psychedelics. Preliminary findings suggest that anesthesia shifts the balance towards segregation, while psychedelics promote integration. Finally, I explore the causal role of the thalamus in modulating this balance using transcranial low-intensity focused ultrasound combined with functional MRI and ISD analysis. This multifaceted approach provides novel insights into the neural mechanisms underlying consciousness, with potential implications for anesthesia monitoring and the treatment of disorders of consciousness.

Pattern and Process Across Temporal Scales of Biodiversity Turnover
Matheus Januario Lopes de Sousa, Ecology and Evolutionary Biology

Across the vastest scales of space and time, biodiversity reveals staggering variability, with species numbers differing dramatically among otherwise similar groups of organisms. Previous studies have shown that species traits, such as physical characteristics or ecological roles, often fail to explain these patterns. This dissertation explores whether emergent, population-level processes influence species formation and extinction rates, thus addressing a core question in evolutionary biology: how and why did life become so diverse? Specifically, I investigate whether sexual selection intensity (Chapter 1) and population-level persistence over decadal timescales (Chapter 2) or millennia (Chapter 3) drive diversification rates at million-year timescales. I also show how mixing different levels of organization, like species and populations, can severely bias inferences on evolutionary rates across millions of years (Chapter 4). Finally, I present an open-source tool for teaching evolutionary biology through an interactive, data-driven framework (Chapter 5). I conclude that a comprehensive understanding of Earth’s biodiversity dynamics demands the integration of evidence across a wide array of disciplines.

Spindle Assembly Checkpoint Dysregulation and Cancer Proliferation
Soubhagyalaxmi Jema, Program in Biomedical Sciences, Barbour Scholar

The spindle assembly checkpoint (SAC) ensures accurate chromosome segregation by halting cell cycle progression until all chromosomes are correctly attached to the mitotic spindle. SAC dysregulation is linked to aneuploidy, disrupting the expression of cell cycle and cancer driver genes and potentially creating a feedback loop that promotes neoplastic growth. Understanding how cancer cells modulate the SAC to drive proliferation remains unresolved. The SAC is activated at unattached kinetochores, where BUB1 recruits and initiates the Mitotic Checkpoint Complex (MCC), which delays anaphase onset to ensure accurate segregation. For accurate chromosome segregation, even a single unattached chromosome must delay anaphase onset by generating MCC at a high rate. Balanced MCC generation is vital for optimal SAC strength and minimizing missegregation. This study investigates SAC dysfunction’s role in aberrant cell division and aneuploidy, enhancing our understanding of its regulation in cancer progression.

Bayesian Modeling and Inference for Complex Multimodal Biological Networks
Guoxuan Ma, Biostatistics

Advancing the integration and analysis of multimodal biological networks is vital for uncovering key mechanisms underlying health and disease. This dissertation develops novel Bayesian methods to address challenges in integrating high-dimensional and multimodal network data in neuroimaging, metabolomics, and microbiomics. The first project, Bayesian Image-on-Image Regression via Deep Kernel Learning (BIRD-GP), introduces a Gaussian Process-based framework for image-on-image regression, integrating multimodal neuroimaging data while ensuring interpretability and inference reliability. The second project, Bayesian Analysis for Untargeted Metabolomics (BAUM), models untargeted metabolomics data with feature-metabolite matching uncertainty, providing robust metabolite selection and pathway analysis. The third project extends BAUM to integrate microbiome data and metabolomics data. These methods achieve state-of-the-art performance while offering biologically meaningful insights. Together, this research advances statistical tools for multimodal biomedical network data, providing pathways to explore complex biological systems and improve health outcomes.

Regulation of Sex Differences by the Ancestral X-Y Chromosomal Gene Pair Kdm5c-Kdm5d
Rebecca Malcore, Genetics and Genomics

Conventionally, the Y chromosome is thought to drive sex differences by triggering differential hormone production. However, differences between the sexes can arise independently of hormones. Here, I show that the ancestral X-Y chromosomal gene pair Kdm5c-Kdm5d regulates sex-biased gene expression in pluripotent mouse embryonic stem cells (ESCs). By RNA sequencing, I show that wild-type (WT) XX female ESCs exhibit >2-fold higher expression of 409 genes compared to WT XY male ESCs. Conversely, WT XY male ESCs exhibit >2-fold higher expression of 126 genes compared to WT XX female ESCs. Loss of Kdm5c in female ESCs downregulates female-biased genes. In contrast, loss of either Kdm5c or Kdm5d in male ESCs upregulates female-biased genes and downregulates male-biased genes, effectively neutralizing sex-biased gene expression. Moreover, I show that ectopic Kdm5d expression is sufficient to drive male-biased gene expression in female ESCs. This work establishes Kdm5c-Kdm5d as critical regulators of sex-biased gene expression.

Mutant Isocitrate Dehydrogenase 1 (mIDH1) in Glioma Cells Causes a Reduction in Adenosinergic Pathway Signaling within the Glioma TME
Brandon McClellan, Immunology

Gliomas are cancers which arise in the central nervous system, primarily in the brain. The glioma cell mutation of isocitrate dehydrogenase 1 (IDH1) has been shown to increase the immune response to glioma by reducing the expression of immunosuppressive molecules. One immunosuppressive molecule, adenosine, is generated by the adenosinergic pathway (AP) and results in a reduced ability of immune cells to kill tumor cells. Based on preliminary data of mutant IDH1’s effects on the tumor microenvironment, my thesis work investigates the hypothesis that mIDH1 alters the immune response by a reduction of AP signaling. Using human and mouse glioma cells in vitro along with in vivo glioma mouse models, I have identified mIDH1-induced changes in the AP in gliomas. Uncovering these differences in the AP provides novel insights on the use of AP inhibitors for glioma treatment based on the isocitrate dehydrogenase mutation status.

The Identification of Key Signals for the Induction of Durable Humoral Immunity
Alexander Meyer, Pharmaceutical Sciences

Despite their clinical success, vaccines suffer from inconsistencies in protective durability, generating antibody responses that last from months to decades among different vaccine products. To identify how to reliably induce long-lasting antibodies through vaccination, I am using a novel vaccine platform called synthetic virus-like structures (SVLS). SVLS mimic the biochemical and biophysical features of enveloped viruses in a modular scaffold, allowing us to distinguish how vaccine components act alone and cooperatively to influence immune responses. I am using SVLS to investigate how several signals—including dose, epitope density, innate immune activation, and CD4+ T cell recruitment—influence the magnitude and durability of antibody responses in mice and nonhuman primates. To support these studies, I have also developed a novel antibody quantitation assay that improves upon traditional techniques like ELISA. Combined, this work will determine the minimal vaccine ingredients needed for durable protection and establish improved tools to measure antibody responses.

Exploring Shared Mechanisms of Nociplastic Pain and Altered States of Consciousness
Niloufar Pouyan, Neuroscience, Barbour Scholar

Nociplastic pain, a chronic condition originating in the brain rather than peripheral organ damage, is characterized by altered sensory processing, suggesting parallels with altered states of consciousness (ASC). This thesis investigates the neurobiological mechanisms of nociplastic pain by comparing it to sensory processing changes induced by nitrous oxide, a non-classical psychedelic known to elicit ASC. Aim one examines nitrous oxide’s effects on sensory processing during a visual task in healthy volunteers, focusing on network-level changes. Aim two conceptualizes nociplastic pain as an ASC, building on recent findings from our group that identify multisensory alterations as a core mechanism underlying nociplastic pain. By contrasting sensory processing alterations induced by nitrous oxide with those observed in nociplastic pain, this work seeks to uncover shared and distinct neurobiological mechanisms, offering novel insights into the interplay between sensory alterations, consciousness, and nociplastic pain.

Linking Macroecology and Physiology: Mechanistic Underpinnings of Widespread Organismal Responses to Global Change
Charlotte Probst, Environment and Sustainability

As climate change transforms temperature regimes across the globe, classic questions in ecology and evolution regarding the influence of temperature on life history traits are becoming increasingly relevant as frameworks to understand and predict the consequences of climate change for biodiversity. In this dissertation, I draw on macroecological and physiological perspectives to understand how species respond to environmental variability both spatially and in the face of global change. Specifically, I examine the cases of warming-linked body size declines and the influence of migration-modified annual temperature regimes on thermal tolerances in the boreal forest. My results will shed light on the mechanisms underlying size declines and assess relative vulnerability to heat stress in one of the world’s most important regions for bird conservation.

Exploring the Molecular Mechanisms of Host-Microbe Symbiosis in Environmentally Ubiquitous Dairy Bacteria
Keerthikka Ravi, Molecular, Cellular and Developmental Biology

My dissertation research explores the host-microbe interactions of two dairy-associated bacteria: Pseudomonas lundensis and Lactobacillus johnsonii. Comparative genomics and transcriptomic analysis revealed the existence of a Ysc-family Type III secretion system (T3SS) in the environmentally ubiquitous cold-storage dairy spoilage bacterium P. lundensis and identified that the expression of this T3SS was both temperature-dependent and host-dependent. This bacterium does not induce disease in mice or humans, indicating a previously unrecognized role for this “virulence” factor. Following the same conceptual framework, comparative genomic and transcriptomic analysis identified host-species specific genes in the ubiquitous beneficial symbiont L. johnsonii, including those encoding cell surface proteins that are uniquely regulated in the host. To further explore their functions, I developed a genetically tractable system for L. johnsonii (using a counter-selectable marker) and validated the approach using a pathway for milk disaccharide utilization.

Enhancing Regulatory T Cell (Treg) Therapies for the Treatment of Autoimmune Diseases
Claire Shudde, Pharmacology

Regulatory T cells (Tregs) prevent autoimmune diseases (ADs) by suppressing self-reactive immune cells that attack and damage healthy tissues. Treg therapies hold great potential for AD treatment, but are hampered clinically by both poor engraftment of Tregs upon transfer into patients and loss of Treg suppressive ability. Because Tregs require IL-2 cytokine-induced JAK-STAT5 signaling to persist and remain functional, we developed a STAT5 activator that can be incorporated into Tregs, supplementing IL-2 treatment. My thesis project has determined that this STAT5 activator improves Treg survival and engraftment into mouse models. When IL-2 is removed, the STAT5 activator can maintain Treg expression of FOXP3, a master transcription factor important for Treg function. Ongoing work investigating the effect of STAT5 activator Tregs in a disease model of induced colitis shows promising initial results. My thesis project has identified a novel strategy to address barriers to the treatment of ADs by Treg therapies.

Oxygen-Dependent Regulation of Tendon Extracellular Matrix
Stephanie Steltzer, Molecular and Integrative Physiology

Tendons are soft, connective tissues that provide stability and movement of joints. Tendons are susceptible to injury, such as rupture and tendinopathy, and there exists few to no therapeutic strategies or surgical approaches available to restore their structure and function. Prior work in tendon healing using mouse models has demonstrated that neonatal, but not adult, tendon can fully restore its structure and function, and that the healing tendon is oxygen deprived. The inability to restore adult tendon function is largely due to the limited capacity for tendon fibroblasts (TFs) to regenerate or remodel their surrounding extracellular matrix (ECM).

Our preliminary findings show that the embryonic Achilles tendon in mice is hypoxic, suggesting low oxygen availability drives ECM deposition by TFs in the developing tendon. Postnatally, tendon structure and function rely on expression of hypoxia inducible factor-1a (HIF1a) by TFs. Yet, overexpression of HIF1a has been observed in biopsies of patients with tendinopathy, introducing a gap in our understanding of the role of HIF1a and hypoxia in tendon development and disease. The overarching objective of our work is to identify the response of TFs to oxygen availability. The central hypothesis is that hypoxia induces TF cell proliferation, and the generation of ECM deposition required for tendon function. To test this hypothesis, we use TFs in culture from wild-type and established HIF1a knockout and overexpression transgenic murine models available in our laboratory.

To measure the temporal response of TFs and their dependence on HIF1a for proliferation, survival, and nascent ECM production in hypoxia we use an optimized three-dimensional (3-D) model for long term in vitro culture. The 3-D culture system provides a physiologically relevant model to investigate micro-tendon structure and function, including for the visualization and alignment of ECM and TF force production in situ. We have seen that hypoxia may induce a decrease in nascent ECM deposition by TFs in addition to increased cell proliferation and survival. We will continue to use multi-omics approaches, including bulk RNA-sequencing and proteomics of the soluble ECM fractions, to discover new pathways for future druggable targets.

Further, using FDA-approved drugs to induce pseudo-hypoxia (e.g., roxadustat) or induce pseudo-normoxia (e.g., belzutifan) of TFs in 3-D tissue culture, we have also found these pseudo-hypoxic environments may reduce nascent ECM and viability while increasing proliferation. Future studies will also test potential new drug candidates from the RNA-sequencing and proteomics findings. Collectively, these findings will identify key mechanisms responsible for the response of TFs to hypoxia that affects ECM deposition and micro-tendon function. This work will also contribute to the discovery of novel targets of hypoxia and HIF1a, with high potential for identifying druggable strategies for improving tendon health.

Dynamics of Body Form Evolution in Lizards and Snakes
Natasha Stepanova, Ecology and Evolutionary Biology

Groups of organisms vary in morphological disparity, with some groups radiating into many different body forms and others showing conserved forms. Explaining this pattern is a key goal of evolutionary biology. I used morphological measurements of lizards and snakes to study how intrinsic properties may facilitate or constrain evolution across groups. First, I studied how evolutionary covariance between traits shifts between two radiations of lizards. I found that a morphologically conserved group has a distinct covariance structure, and that a punctuated model best fits the data. I then tested whether decoupling of the limbs has facilitated the reduction and loss of limbs in some groups. Using a large dataset of vertebral counts I collected, I also asked how scaling relationships change across groups and whether within-species variance impacts capacity to evolve and disparity between groups.

Strategic Catalytic Reactions for the Synthesis of Biologically Active Chemical Motifs
Austin Ventura, Medicinal Chemistry

Alkanes/alkenes, aldehydes, alkyl amines and ethers are abundant chemical feedstocks that are often used as raw materials in the chemical and pharmaceutical industry. Existing methodologies for metal-catalyzed cross-couplings of these commodity chemicals rely on preinstallation of reactive functional groups on both substrates. In contrast, C–H functionalization approaches offer promise in simplification of the requisite substrates and have recently been developed to afford value-added C(sp2/3)−C bonds that are useful for drug development and late-stage modification of peptides; however, challenges from low or similar reactivity of various C–H bonds introduces considerable complexity. This work describes the optimization and initial mechanistic investigations of three Ni-catalyzed C–C bond forming reactions. In total, these reactions showcase the utility and compatibility of nickel catalysis under thermal redox conditions to provide value-added C−C bonds that are important to both medicinal and synthetic chemists.

A Deep Learning Approach to Enhancing Maternal and Fetal Health Outcomes of Preeclampsia using Electronic Health Records and Genetic Data
Xiaotong Yang, Computational Neuroscience

Preeclampsia (PE), a pregnancy complication marked by high blood pressure and protein in the urine, affects approximately three to nine percent of pregnancies worldwide and three percent in the US. PE is a major cause of maternal and fetal mortality and morbidities, with long-lasting effects on both the mother and child’s well-being, increasing the risk of future cardiovascular disease and metabolic disorders in mothers and developmental disorders in offspring.

This thesis aims to improve clinical decision-making for PE patients using advanced deep learning models. First, it predicts delivery timing for PE patients based on electronic health records (EHR). Second, it jointly estimates the onset of maternal adverse effects post-PE with a novel survival prediction model. Third, it explores drug repurposing strategies tailored to specific PE subtypes for prevention. By leveraging machine learning on EHR data, this research seeks to enhance personalized care, optimize treatment timing, and identify potential therapeutic options, ultimately improving maternal and fetal outcomes in PE cases.

Supervised Feature Discovery and Uncertainty Quantification in High-Dimensional Regression
Leyao Zhang, Biostatistics

Fusing weak regression predictors into homogeneous groups can create strong predictors with shared group-level effect sizes, reducing parameter space dimensions and enhancing statistical power. Chapters I and II introduce a novel supervised learning approach employing L0-fusion, solved via mixed integer optimization (MIO), to cluster features in linear regression and time-series models, enabling simultaneous feature selection and estimation. Chapter I focuses on identifying key items from a quality of life questionnaire that are strongly associated with clinical outcomes, such as lung function and asthma control, creating adaptive summary metrics that improve statistical power and clinical interpretability. Chapter III applies an MIO-based approach to identify sewage sampling manholes most predictive to COVID-19 cases, optimizing community-level disease surveillance and cost-effectiveness. To address challenges in uncertainty quantification when fusion and estimation are combined, Chapter III introduces ensemble inference, a novel framework addressing post-fusion uncertainty quantification, constructing valid confidence sets through adversarial noise perturbation.

Engineering Systemic STING-Activating Metalloimmunotherapy
Xingwu Zhou, Pharmaceutical Sciences

Natural systems utilize metal ions in structurally ordered architectures to regulate essential biological processes, inspiring the rational design of nanostructures. The cGAS-STING pathway detects cytoplasmic DNA and triggers type I interferon (IFN-I) responses essential for antitumor immunity. However, translating STING activation into effective cancer immunotherapy has been challenging due to poor pharmacological properties. Here, I developed CRYSTAL (Crystal-like STING-Activating nanoassemblies), a structurally ordered intermetallic nanoparticle for potent systemic STING activation. CRYSTAL was synthesized via self-assembly of manganese (Mn²⁺) ions intercalated with cyclic-dinucleotides, wrapped within lipid layers. Computational modeling accurately predicted CRYSTAL’s self-assembly behavior, highlighting its tunability and design precision. At ultra-low doses, intravenous CRYSTAL administration effectively activated STING in mice, dogs, and non-human primates, and CRYSTAL demonstrated remarkable antitumor efficacy in late-stage murine and rabbit tumor models. CRYSTAL also induced IFN-I responses in human head and neck squamous cell carcinoma biopsies, underscoring its translational potential. These findings establish structurally ordered metallo-nanostructures as a promising strategy for metalloimmunotherapy.