Alliance One Tobacco (Malawi)
Academic Research Repository at the Institute of Developing EconomiesNot a member yet
52305 research outputs found
Sort by
Robust Uncertainty Quantification for Non-Negative Matrix Factorization with Applications to Mutational Signatures Analysis
No full textMutational signatures are distinctive patterns of mutations resulting from cancer-causing molecular processes, such as defective DNA repair mechanisms or UV radiation. The study of mutational signatures in cancer has provided insights into the molecular epidemiology of cancer and helped guide clinical decisions and therapies. Non-negative matrix factorization (NMF) methods have been foundational to the discovery of many mutational signatures and their respective loading or activity level in individual tumor genomes. In this dissertation, we explore topics related to robust uncertainty quantification for NMF methods with applications to mutational signatures analysis in cancer. Chapter 1 provides an overview of current literature and key concepts. Chapters 2 and 4 introduce and characterize specific models for mutational signatures analysis, while Chapter 3 addresses a general statistical challenge in developing and testing robust NMF models.
In Chapter 2, we introduce BayesPowerNMF, a Bayesian NMF method with uncertainty quantification for mutational signatures analysis that is robust to model misspecification. While existing NMF models have been successful in discovering many mutational signatures with verified etiologies, the NMF model is ultimately only a rough approximation to reality. Model misspecification, or using a model that deviates from reality, can lead to poor inference, like failing to detect important mutational processes or inferring spurious ones that do not actually exist. In BayesPowerNMF, we leverage power posteriors for nonparametric robustness to misspecification. By performing full Bayesian inference, we are able to report uncertainty in both the signatures and loadings inferences. In simulations of both well-specified and plausibly misspecified genomic data, we illustrate the limitations of two leading NMF methods for mutational signatures discovery and demonstrate that BayesPowerNMF discovers more true processes and fewer spurious processes than these leading NMF models. Finally, we demonstrate BayesPowerNMF's performance on whole-genome sequencing data from the ICGC/TCGA Pan-Cancer Analysis of Whole Genomes project.
In Chapter 3, we formulate a maximum density method for specifying the parameters of Dirichlet distributions that provides better control over the location and scale of the density near the boundary of the simplex than conventional approaches. Dirichlet distributions are widely used in Bayesian models and are a natural choice when modeling mutational signatures, such as for formulating informative priors based on known signatures for Bayesian NMF models and generating realistic mutational signatures for simulation studies. In the maximum density method, we tune the parameters to maximize the Dirichlet density at a specified target location point, subject to a scale constraint. The scale constraint is very flexible: for instance, for modeling mutational signatures we constrain an approximation to the cosine error, which is the preferred similarity metric in this setting. We demonstrate several desirable features of our maximum density method in a series of examples, including defining Metropolis--Hastings proposals for MCMC, constructing prior distributions for rare events, and generating simulated probability vectors for mutational signatures analysis.
In Chapter 4, we propose methods for uncertainty quantification for non-negative least squares (NNLS) regression. NNLS is a popular method for loadings-only inference in mutational signatures analysis, where we wish to estimate the activity of a fixed set of signatures in a single tumor genome. These tools can be used to estimate the activity of known mutational processes in tumor genomes to guide clinical treatment and validate biomarker mutational signatures discovered via NMF studies. While there is extensive work in loadings-only NMF inference methods to improve point estimates, these methods lack uncertainty quantification. We consider two approaches to building uncertainty quantification for NNLS regression: first, by leveraging the equivalency between NNLS and ordinary least squares linear regression for non-zero loadings to build confidence intervals, and, second, by resampling the observed mutation counts vector and repeating NNLS on each replicate to build bootstrap confidence intervals. In simulation studies, our resampled NNLS method performs well in both estimating the loadings and classifying active vs inactive signatures, with interpretable uncertainty quantification for both tasks and well-calibrated confidence intervals for loadings estimates.Biostatistic
Expanding the toolbox for deubiquitylase pharmacology
No full textUbiquitylation is a posttranslational modification that governs protein cellular fate. Balanced ubiquitin signaling is vital for regulating cell function, replication, and survival, and its dysregulation is implicated in various indications such as neurodegeneration, cancer, autoimmunity, and neurodevelopmental disorders. Ubiquitin is covalently attached to substrates in a stepwise manner by an E1, E2, E3 enzymatic cascade, and is removed by a class of isopeptidases called deubiquitylases (DUBs). DUBs can stabilize both oncogenic drivers and tumor suppressors in cancer. Additionally, both loss-of-function and gain-of-function DUB mutations can cause or drive disease pathologies. Therefore, targeting DUBs with small-molecule tools offers direct and indirect therapeutic approaches for disease intervention.
The work in this thesis expands the available chemical toolbox for the interrogation of DUB biology in healthy and disease states, and provides starting points for DUB-targeted therapeutic development. I focus on two DUBs, USP7 and USP48. The potential for USP7-targeted therapeutics in cancer via the regulation of the p53/MDM2 pathway is well-documented. More recently, USP7 has emerged as a critical regulator in neurodevelopment, as mutations in the USP7 gene were identified as causal for the rare neurodevelopmental disorder, Hao-Fountain syndrome (HAFOUS). Our interest in USP48 arose from recent studies by Dana-Farber colleagues that identified USP48 knockout as a novel sensitizer to hypomethylating agents in acute myeloid leukemia and potentially other cancer types.
In Chapter 2, I describe the discovery and rigorous mechanistic characterization of MS-8, a first-in-class small-molecule DUB activator. I demonstrate that MS-8 allosterically engages ubiquitin-specific protease 7 (USP7) and induces enzymatic activation, similar to the mechanism of auto-activation by the C-terminal tail of USP7. We assess the therapeutic potential of our activator in vitro and in cells against a panel of HAFOUS patient-derived USP7 variants and identify a subset of mutations sensitive to MS-8 driven activation. Additionally, we provide an initial structure-activity relationship for MS-8 and highlight viable exit vectors on the scaffold for the development of heterobifunctional DUB recruiters. In Chapter 3, I introduce a next generation covalent inhibitor for USP7. Using a structure-guided approach, we elaborate the reported noncovalent Compound 41 into a covalent USP7 inhibitor. Through a focused medicinal chemistry campaign, we optimize our covalent series and validate its useability as a cellular probe in p53-wildtype and p53-mutant cell lines. Lastly, in Chapter 4, I describe the identification and validation of first-in-class selective inhibitors of USP48. I performed a USP48-focused high throughput screen then triaged screening actives in a series of orthogonal assays and counter screens to credential 3 novel scaffolds. We demonstrate the chemical tractability of one of our hits, MD-156, to support further medicinal elaboration. In this chapter, I additionally explore USP48 intramolecular regulation by its C-terminal ubiquitin-like (UBL) domain and define the consequences of UBL domain deletion to USP48 enzymatic activity and substrate-recognition.
Ultimately this thesis provides a novel USP7 activator and a framework for future DUB activator discovery, as well as two inhibitors for USP7 and USP48 with pharmacological potential for the treatment of cancers. These agents are important tools for studying the detailed mechanism of pathologies of the DUBs in their respective diseases. Moreover, these compounds and associated knowledgebases set the stage for future DUB-focused biological discovery and targeted therapies.Chemical Biolog
Statistical Methods for Negative-Unlabeled Data with Application to Long COVID
No full textNegative-unlabeled data arise in settings where a subset of observations has known negative outcomes (e.g., people without a history of SARS-CoV-2 infection do not have Long COVID), while the remaining observations are unlabeled (e.g., individuals with prior infection have uncertain Long COVID status). This partial labeling structure presents challenges for statistical inference, particularly in characterizing heterogeneous conditions such as Long COVID (LC). Existing supervised, unsupervised, or semi-supervised approaches cannot directly accommodate negative-unlabeled data. However, data with this structure are increasingly common in public health, including electronic health records and post-infectious syndrome research. In this dissertation, we develop statistical approaches tailored for negative-unlabeled data with applications to LC and potential extensions to other partially labeled disease phenotyping problems.
LC is a multisystem condition with variable pathophysiological manifestations. Its fluctuating symptom patterns can persist for months or years following acute SARS-CoV-2 infection. Understanding its clinical heterogeneity and longitudinal trajectories remains an urgent research priority. This work is motivated by the NIH-sponsored Researching COVID to Enhance Recovery (RECOVER) Initiative, a large observational cohort study of individuals followed quarterly over multiple years. The RECOVER study features negative-unlabeled outcomes, as it includes participants without a history of SARS-CoV-2 infection. These uninfected individuals establish the baseline symptom prevalence and variability in the general population and can thus serve as a valuable control group for studying LC, provided that the statistical model is capable of appropriately accommodating negative-unlabeled data. The combination of partial labeling of LC status, high-dimensional heterogeneous symptom data, and irregular follow-up schedules in RECOVER presents significant analytic challenges that need to be addressed by the statistical methods proposed in this dissertation.
Chapter 1 proposes a sparse Bernoulli mixture model (BMM) with novel parameterization to achieve feature selection. The method can identify symptom-based latent clusters and select a minimally informative set of features. We extend the proposed BMM to leverage negative-unlabeled data to improve the cluster identification. An application to the RECOVER-Adult Cohort data reveals 3 LC indeterminate clusters and 2 LC subphenotypes, as well as 11 crucial symptoms for defining LC.
Chapter 2 extends this framework to longitudinal data by developing a latent Markov model (LMM) with structured transition matrices. The proposed LMM accommodates negative-unlabeled outcomes by incorporating data from uninfected individuals, yielding improved performance compared to approaches that rely solely on infected participants. In addition, the model naturally handles sparse longitudinal data arising from missed visits and staggered enrollment. Chapter 3 applies the proposed LMM to the RECOVER-Adult Cohort data and further adapts the model to include clinical covariates. The method uncovers 5 latent clusters and identified key factors associated with LC persistence or recovery. Collectively, these methods provide a flexible and robust analytic framework for negative-unlabeled data, with applications both within and beyond the context of LC.Biostatistic
Dissecting spatial tumor-immune microenvironment in response and resistance to immune checkpoint blockade in metastatic melanoma
No full textImmune checkpoint blockade (ICB) therapies have markedly improved the prognosis for patients with stage III & IV metastatic melanoma by prolonging progression-free and overall survival rates. However, the variability in mechanisms of immune evasion and resistance present significant challenges in the clinical efficacy of ICBs. This project aims to define drivers of immunotherapy response and resistance by employing advanced genomic, single cell mRNA analyses, and spatial profiling techniques on tissue biopsies from metastatic melanoma patients.
In this study, we developed a framework to analyze response and resistance, both intrinsic and acquired, via immune features in the tumor microenvironment in a standardized, uniformly processed, and deeply clinically annotated cohort of metastatic melanoma patients (n=61) treated with ICB as part of the human tumor atlas network (HTAN) initiative1,2. From the tumor samples, we conducted single-nucleus RNA sequencing, and for a subset of the samples, high-resolution spatial imaging (including protein mIHC, CODEX, and MERFISH transcriptomics). Standardized processing and data pipelines allowed for the integration of genomic, transcriptomic, and spatial features to elucidate characteristics and mechanisms in the tumor microenvironment and their relationships with resistance. For this thesis, I focused on MERFISH spatial transcriptomics analysis.
Single-nucleus RNA sequencing analysis revealed CXCL13+CD4+ T cells and ISG+CD8+ T cells as the strongest predictors of durable clinical benefit (DCB) among all immune populations, independent of clinical confounders. We also identified five recurrent cellular neighborhood modules: extreme responders are enriched in B cell–enriched RCNs with close T–B cell distance, non-responders in tumor–myeloid and tumor–myeloid–stromal interface RCNs, and non-extreme responders in tumor-dominant and stromal–immune interface RCNs. This project integrates transcriptomic and spatial features to elucidate shared tumor and microenvironmental states and their relationships with resistance, guiding more personalized and effective treatment strategies for metastatic melanoma.Graduate Educatio
Civic Discourse Pedagogies in American Higher Education: Tracing Historical Patterns, Enduring Paradigms, and Future Possibilities
No full textCivic discourse on college campuses has become a flashpoint in American politics. Amid deep political division, rising threats to freedom of inquiry and speech, and declining public trust, colleges and universities are seeking ways to teach and practice the communication across difference so essential to their civic, educational, and scholarly missions. Importantly, this is not the first time institutions of higher education have sought new approaches to civic discourse education in response to profound societal changes and challenges. They have done so repeatedly, stretching back to higher education’s earliest days in colonial America. In this dissertation, I draw on histories of American politics, higher education, ideas, and philosophy to make two core arguments. First, I demonstrate that the emergence of civic discourse pedagogies in American higher education has followed a historical pattern: Societal change has generated new visions of politics; together with the transformative forces that shaped them, these visions have led to new conceptions of higher education’s mission and new ideas about the kinds of citizens higher education should aim to produce. These conceptions have demanded new approaches to civic education, spurring the creation of new ways to teach civic discourse. Second, I argue that these new pedagogies were not just instructional innovations. Rather, they were products of new paradigms of civic discourse practice and pedagogy, rooted in specific notions of personhood, power, and social epistemology. I build this argument by analyzing three examples of pedagogical emergence: the development of forensic debate in the mid-1700s; deliberative discussion in the early 20th century; and intergroup dialogue in the late 1980s. These models of pedagogical emergence and paradigm development constitute new scholarly contributions—ones that provide analytical tools that can help us understand and reframe our contemporary challenges of civic discourse. In my final chapters, I demonstrate this point by applying these models to our recent history. Ultimately, my hope is that viewing campus civic discourse in historical perspective can support the vital work of sustaining conversations across difference amid the critical challenges facing higher education today.Educatio
Stromal-immune cell crosstalk in the developing brain borders
No full textCentral nervous system (CNS) border tissues including the meninges serve as key immune interfaces between the brain and the periphery. The three meningeal layers – the pia mater, the arachnoid mater, and the dura mater - each host a unique landscape of immune lineage cells that continuously shape healthy adult brain function and provide a firewall against invasion of pathogens into the central nervous system. However, much less is known about how the meningeal immune compartment develops in early life or how meningeal immune cells may interact with or protect the brain during critical developmental windows. Emerging evidence also suggests that meningeal stromal cells critically shape the adult meningeal immune landscape. How meningeal stromal cells contribute to establishment of this compartment in the early life window is an outstanding question.
To better understand the how the meningeal immune compartment is established, we profiled changes in the composition of meningeal immune cells from early postnatal life through adulthood. Unexpectedly, we uncovered a role for the meninges as a niche for B lymphopoiesis in the early postnatal window. B cells appear in the dura mater and other meningeal compartments as a wave spanning the first month of life in mice. Meningeal B cells undergo lymphopoiesis locally but in concert with other waves of extramedullary lymphopoiesis across the body that fuel generation of the B2 B cell compartment. Developing B cells in the dura are seeded by a common pool of hematopoietic progenitors as are B cells in other early life organs but diverge in the perinatal window and establish local hematopoietic foci in microanatomical niches near the dural venous sinuses. To elucidate the niches that support meningeal lymphopoiesis and their relationship to bone marrow niches, we profiled stromal and endothelial cells from the early life dura mater and bone marrow. This data revealed that dura and bone marrow niches are fundamentally distinct in composition. Indeed, the early life dura mater lacks canonical mesenchymal stromal cell populations known to regulate B cell development and instead comprises a heterogenous landscape of fibroblast-like cells (FLCs) enriched for the transcription factor Foxd1. Lymphopoietic foci associate with a unique network of dural sinus-associated (peri-sinus) FLCs that express the pro-hematopoietic chemokine Cxcl12, and ablation of Cxcl12 from Foxd1-expressing FLCs profoundly impairs local B lymphopoiesis. These data suggest that CNS border tissues contribute to generation of the postnatal B cell compartment and provide a model for how extramedullary stromal niches may regulate lymphopoiesis across the body in early life.
Separately, we have undertaken broader efforts to define functional roles for CNS immune cells in health and disease. One of these includes a collaborative effort to develop FEAST (Flow cytometric Engulfment Assay for Specific Target proteins) - a robust, in vivo assay to quantify engulfment of neuronal and myelin substrates by brain and brain border macrophages. This approach will be useful for studying the role of CNS phagocytes in tissue development, infection, and other neurological diseases across the lifespan.Immunolog
Process as Truth: On the Career of Tahara Sōichirō and a Culture of Mass Media Critique in 1970s–2000s Japan
No full textJapanese politics went “live” in the 1990s, a turbulent period marked by the collapse of the bubble economy and growing public dissatisfaction with endemic corruption in the ruling Liberal Democratic Party. Criticized for their lack of transparency, politicians found themselves navigating a new and chaotic arena of Sunday morning television debate shows that effectively bypassed the closed channels by which they had historically communicated with journalists. At the forefront of this transformation was journalist and commentator Tahara Sōichirō, whose debate programs Asa made nama terebi! and Sunday Project pioneered the genre. Tahara described his mission at the time as unveiling the previously hidden “process” of politics to the Japanese public through candid debate. The time had come to see inside politics and media.
Critiques of mass media’s relationship with politics, however, e.g., that journalists were embedded in overly cozy relations with politicians—making critical reporting difficult, if not impossible—and that the national newspapers and the semi-public broadcaster NHK’s commitment to “objectivity” or “neutrality” made their coverage deeply unsatisfying, were decades old by the 1990s. Thus, to understand the appeal of seeing politics unfold in a less controlled media environment, it is essential to examine a longer history of journalism in Japan as practiced outside the dominant national newspapers and NHK. This dissertation constructs a genealogy of such “outsider” journalism in magazines and commercial television in 1970s–2000s Japan, one that positioned itself as critical of Japan’s closed mass media structure. At the heart of this journalism was the practice of “media reflexivity,” which I define as the act of acknowledging one’s position within the media to encourage the viewer or reader in turn to consider their own existence inside of media. This reflexivity implicitly posed several radical questions: Why was the established mass media unwilling or unable to communicate knowledge about the process by which it covered or created events? Why could it not speak about itself? I argue that such media reflexivity, as it crystallized in the 1990s debate programs, paralleled the development of “reflexive modernity” in Japan—marked by growing social literacy and precarity in response to economic stagnation and the hollowing-out of postwar social structures.
This dissertation traces Tahara Sōichirō’s career as it developed across four worlds of journalism. Chapter 1 examines 1970s investigative journalism, catalyzed by journalist Tachibana Takashi’s iconic 1974 exposé of LDP leader Tanaka Kakuei in Bungei shunjū. Chapter 2, in turn, assesses the magazine Uwasa no shinsō in the 1980s and the genre of mass media criticism (masukomi hihyō)—a group of publications that sought to reflexively document contemporary media saturation. Chapter 3 explores the influence of television station TV Asahi in pioneering a new form of opinionated “television journalism” in the late 1980s, and the related cultural figure of the anchorman or “caster.” Finally, Chapter 4 and the Conclusion analyze the rise and decline of so-called “telepolitics” in 1990s Japan, assessing how debate programs such as Tahara’s Sunday Project destabilized Japanese politics by injecting spontaneity into political communication. These programs’ influence, however, waned during the Koizumi Junichirō administration (2001–2006), which emphasized Koizumi’s impromptu and “authentic” communication style, thereby allowing the LDP to co-opt many of the criticisms made of its historical relationship with the media. Tahara Sōichirō closely backed Koizumi and served as a media fixer in this transitional period. The dissertation closes with a brief discussion of the legacy of telepolitics in post-Koizumi-era Japan.East Asian Languages and Civilization
Arsenic Exposure and the Epigenome: Evaluating Effects of Periconceptional and Gestational Arsenic Exposure on DNA Methylation
No full textExposures before conception can impact sperm and oocyte epigenomes, morphology, and function with potential consequences for the developing fetus and child. The motivation of this work is to evaluate how early life exposure to arsenic can impact offspring epigenomes. To address this, we begin by studying the impacts of maternal arsenic levels on DNA methylation and DNA methylation-based age estimations that serve as potential approximations of gestational maturity in cord blood and placental tissues. Our findings indicate that there are sites sensitive to arsenic exposure across populations and tissue types and that biological clocks that estimate gestational duration are sensitive to arsenic exposure. We then explore DNA methylation as a function of parental arsenic exposure in a study of children born with spina bifida to determine whether observations about DNA methylation in this group can shed light on the biological pathways linking arsenic exposure and spina bifida – a neural tube defect with complex etiology. While specific arsenic-associated etiology remains unclear, our results indicate that DNA methylation of tissues taken from children born to mothers with higher arsenic exposure differs from those taken from mothers with lower arsenic. Finally, we evaluate the role of folate and arsenic on rDNA methylation of sperm in mice exposed through their mothers. We conclude from these investigations that arsenic and folate impact rDNA methylation with specific loci in the rDNA being sensitive to variations in both. Taken altogether, this work demonstrates the role of gestational arsenic on the epigenome in multiple tissues and sheds light on this relationship in understudied tissues and segments of the DNA.Population Health Science
Beyond Narrative: The Rise of Historical Scholarship in Song Historiography
No full textThis dissertation explores the development of historiography in the Song dynasty, focusing especially on two distinctive undertakings: the evidential investigation of historical sources and “facts” as Song scholars understood them, and the pursuit of historical analysis, interpretation, and explanation in the form of historical commentary. I argue that whereas earlier historians had treated narrative as the core, if not the sole, task of historical writing, the emergence of new historiographical genres in the Song and the shifts in historical consciousness that accompanied them gave rise to a distinct domain of historical scholarship alongside narrative history. The dissertation seeks to delineate these changes, explain their emergence within broader political, intellectual, and historiographical contexts, trace their subsequent development, and assess their long-term significance for the evolution of Chinese historiography.
Chapters One and Two focus on historical evidential investigation. Chapter One examines Sima Guang 司馬光’s Zizhi tongjian kaoyi 資治通鑒考異—the first work to systematically lay out procedures for comparing, analyzing, and selecting among sources. Beyond reconstructing Sima Guang’s methods of evidential investigation, this chapter also seeks to contextualize the background in which this endeavor took shape and to explain why he was so determined to make it available to a broader public. Chapter Two turns to two Southern Song historians, Li Tao 李燾 and Li Xinchuan 李心傳, situating them within the aims and constraints of contemporary history compilation in the Song to understand and analyze how they inherited and transformed Sima Guang’s evidential investigation.
Chapters Three and Four then turn to historical commentary. Chapter Three first traces changing attitudes toward personalized commentary and interpretation of history and shows that this previously much-contested practice came to be widely accepted and actively practiced in the Song. It then proposes possible explanations for this shift and, finally, highlights the new forms of historical consciousness that emerged in historical commentary. The last chapter offers a detailed reading and analysis of the different interpretative tendencies and explanatory orientations found in these commentaries.East Asian Languages and Civilization
Demonstratives and the semantics–pragmatics interface
No full textOne of the most fundamental properties of language is its ability to refer to entities both in the world and those established in discourse. Across languages, speakers use a range of expressions to introduce, maintain, and distinguish referents. Among these, definite descriptions (for example, the book) and demonstrative descriptions (for example, that book) are two common ones that have been central to research on the semantics and pragmatics of nominal reference. Despite the considerable overlap in their distribution, demonstratives differ from definites in certain key aspects. This dissertation focuses on two cross-linguistically robust properties of demonstratives that distinguish them from definites: (i) their anti-uniqueness effect and (ii) their selective behavior in discourse anaphora. I examine these phenomena as a diagnostic window on the semantics–pragmatics interface, with the aim of clarifying the division of labor between grammatical meaning and discourse-level reasoning in the referential domain.
The first part of the dissertation examines anti-uniqueness, the constraint that makes the use of demonstratives infelicitous in inherently unique contexts (for example, #That sun is bright). Drawing on novel data from Bangla, I show that the anti-uniqueness effect extends beyond demonstratives to certain classes of definites. I argue in favor of the presuppositional approach to anti-uniqueness, illustrating that it is better suited than approaches based on pragmatic economy to account for typological variation in this domain. I then introduce a modalized implementation of the presupposition that offers broader empirical coverage.
The second part turns to discourse anaphora, introducing quantitative diagnostics for how definites and demonstratives are licensed across typologically diverse languages. Controlled experiments across English, Turkish, and Bangla reveal a consistent cross-linguistic asymmetry: demonstratives are uniquely sensitive to subtle information-structural contrasts, unlike definites. This empirical pattern isolates the pragmatic signature of demonstratives and provides a new experimental paradigm for testing competing theories of definite reference, particularly in languages whose definiteness systems are complex or remain under debate, as illustrated for Mandarin and German.
The final part extends this pragmatic paradigm to investigate whether discourse-level constraints on reference are computationally tractable in artificial language systems. Using nineteen large language models, I examine how model size and architecture affect their ability to encode these distinctions. I illustrate that larger models show remarkable sensitivity to the discourse-structural constraints of demonstratives and closely mirror human behavior, whereas smaller models do not. These results provide the first systematic assessment of discourse-pragmatic competence in large language models, showing that the fine-grained pragmatic distinctions underlying demonstrative use are not easily learnable from surface-level statistics alone and emerge only in more advanced models.
In sum, this dissertation integrates theoretical, experimental, and computational approaches to reveal how demonstratives illuminate the interaction between meaning and discourse across natural and artificial language systems. It (i) argues for a dissociation between the anti-uniqueness effects of referential expressions and their deictic component, and illustrates that the anti-uniqueness constraint is best analyzed as a semantically encoded feature, (ii) identifies the pragmatic constraints that govern demonstrative anaphora through cross-linguistic experimentation, establishing a robust empirical diagnostic, and (iii) evaluates how such discourse-level patterns emerge in large language models, providing a computational perspective on the semantics–pragmatics interface.Linguistic