1983 research outputs found
Sort by
A Unified Framework for Collaborative Knowledge Graph Construction, Editing, and Distribution
No full textKnowledge graphs (KGs) have emerged as a critical technology for grounding artificial
intelligence systems in structured facts, offering a solution to the hallucination and relia-
bility issues plaguing large language models (LLMs). Despite their utility, the infrastruc-
ture required to construct, store, version, and collaboratively edit large-scale KGs remains
fragmented. Previous work has addressed individual aspects of graph management but
has failed to provide a unified, version-controlled ecosystem that supports the property-
rich graphs required by modern applications. To address this infrastructure gap, this thesis
introduces a comprehensive framework comprising four integrated systems: Optimus,
a reproducible pipeline for graph construction; Diamond, a novel lossless binary com-
pression format; GitGraph, a semantic version control system; and GraphEnv, an en-
vironment for multi-agent collaboration. We implemented this framework to enable the
end-to-end lifecycle of graph development, from initial data ingestion to downstream appli-
cations. We utilized Optimus to construct OptimusKG, a biomedical KG with 192,307
nodes, 21.5M edges, and 88.6M properties, demonstrating a 56.5% reduction in build time
through parallel execution. To address storage bottlenecks, we developed the Diamond
algorithm, which we benchmarked against standard formats, achieving a 34×compression
ratio on the popular PrimeKG dataset while preserving all node and edge properties. Fur-
thermore, we formalized the theory of graph versioning by developing a three-way merge
algorithm that allows for semantic, structure-aware conflict resolution, enabling true dis-
tributed collaboration. Finally, we integrated these tools into GRENCE, a clinical decision
support application that uses our infrastructure to ground LLM reasoning in verifiable
medical data. This work establishes a robust software engineering foundation for KGs,
transforming them from static artifacts into dynamic, evolving knowledge stores that can
be efficiently maintained by hybrid teams of human experts and autonomous agents.Computer Scienc
Statistical methods for polygenic risk prediction from biobanks and genome-wide association studies
No full textPolygenic risk scores (PRS) have emerged as a promising tool to translate genomic discoveries into clinic decision making. By aggregating the effects of risk-associated genetic variants across the genome into a single number, PRS can quantify a patient’s genetic predisposition for a wide range of health outcomes. The past decade has seen an explosion of statistical methodology to build PRS prediction models, but these existing methods face limitations in prediction accuracy, computational efficiency, and generalizability across populations. In this dissertation, we present three novel approaches to tackle these challenges.
In Chapter 1, we propose ALL-Sum, an ensemble learning-based PRS method that uses summary statistics from genome-wide association studies (GWAS). ALL-Sum leverages L0L2-penalized regression and fast optimization algorithms to enable high prediction accuracy while also dramatically reducing the computational runtime and memory usage. Then, in Chapter 2, we propose SPLENDID, which models gene-by-ancestry interactions to simultaneously capture shared and heterogeneous genetic effects without categorizing individuals into discrete ancestry groups, allowing for fairer clinical implementation in diverse patient populations. Finally, in Chapter 3, we propose STELLAR, which ensembles multiple prediction modeling approaches and functional genomic annotations to flexibly estimate rare variant effects for more comprehensive genome-wide PRS.
Altogether, the work from this dissertation introduces new statistical frameworks to efficiently compute accurate PRS from large-scale genomic data. Each method demonstrates substantial improvements over the current state-of-the-art through comprehensive simulation studies and application to real data. These tools can be used to develop new prediction models for complex traits and diseases and ultimately advance precision medicine.Biostatistic
Laboratory evolution of botulinum neurotoxins for therapeutic applications
No full textThe goal of this thesis is to expand the therapeutic potential of botulinum neurotoxins (BoNTs) by using directed evolution to reprogram both major functional domains. Specifically, we aimed to engineer the protease domain to cleave novel, disease-relevant substrates and to reprogram the receptor-binding domain towards targeting of non-neuronal cell types, such as cancer cells. Through this work, BoNTs are transformed from neurotoxins into programmable delivery systems for targeted therapies, with applications ranging from cancer treatment to broader modulation of cell fate.
Cancers evolve numerous mechanisms to evade both cell death and immune surveillance. Pyroptosis, an inflammatory form of programmed cell death, offers a promising therapeutic strategy by eliminating cancer cells while simultaneously activating immune responses. In this work, phage-assisted evolution was used to reprogram BoNT proteases to cleave two key activators of pyroptosis: procaspase-1 and gasdermin D. A substrate profiling platform was developed to assess the specificity of both wild-type and engineered proteases. While the wild-type protease failed to induce cell death, evolved variants triggered robust cytotoxicity in multiple cancer cell lines. The gasdermin D-cleaving variant induced strictly pyroptotic death, while the procaspase-1-cleaving variant triggered both pyroptosis and apoptosis, suggesting broader caspase-like activity. To enable targeted delivery, evolved proteases were reconstituted into BoNT toxins containing the native translocation domain, resulting in selective death of cancer cells but not non-cancerous cells. These findings demonstrate that BoNT proteases can be reprogrammed to modulate inflammatory cell death and serve as programmable tools for targeted cancer therapy.
In addition to protease engineering, we reprogrammed the BoNT receptor-binding domain as a potential strategy for cell-specific targeting. Using two target nomination strategies—manual selection and a weighted alignment tool—we successfully evolved binders against all nominated targets. A bacterial two-hybrid circuit paired with phage-assisted evolution supported the evolution of binders to CD44v6, CD30, ACHA3, CD1b, RXFP1, and TSN8. Notably, evolved variants targeting ACHA3 and RXFP1 bound to full-length ectodomains with single-digit nanomolar affinity, exceeding the affinity of the wild-type BoNT HC for its native receptor. While in-cell and in vivo validations are ongoing, this work lays the foundation for BoNT HC evolution, analogous to antibody engineering strategies.
Collectively, these studies demonstrate that BoNTs are highly reprogrammable and can be evolved for diverse therapeutic applications. By combining protease and receptor-binding domain engineering, this platform offers a blueprint for developing programmable protein-based therapeutics that modulate cell fate with precision.Chemical Biolog
The Living Fossil Record: Crafting Time and Place through Natural Historical Narratives of Persistence and Extinction
No full textThe middle of the nineteenth century was in many ways the advent of the modern era for natural history as in broader North American culture. Each chapter in this dissertation centers an object of natural historical interest that represented a temporal anomaly or transition point for naturalists caught between eager anticipation of the advances of settler colonial modernity and sadness or unease at the destruction of nature that was narrativized as pre-historic. This ambivalent fascination with species, human groups, and environmental spaces that were made to represent a rapidly disappearing past was a driving force of the cultural reception of natural history during this era. Newfoundland and Labrador as North Atlantic colonial hinterlands served as a focal point for these hopes and anxieties. The effort to cast sea serpents, the Labrador Peninsula, and the great auk and Beothuk as contemporary relicts of a pre-modern age at the Atlantic margins of settler colonial reach was part of a larger effort at scientific storytelling that sought to justify the transformation of the natural world that colonialism had wrought.
This project asks: what does it mean that naturalists in the nineteenth century sought to cast certain contemporary organisms, landscapes, and people as relicts of a prior era? What meaning did the “living fossil,” the “prehistoric landscape” and the “extinct” hold for a public grappling with recent serious shifts in their understanding of the age of the earth and the nature of species within deep time? The answers to this question are varied and complex but ultimately reveal that these temporal narratives were created by natural historians to explain and justify their own place in the contemporary era. The designation of relict or living fossil was a byproduct of the construction of natural historical modernity that was occurring simultaneously. This natural modernity was not defined by any quality of its own, but rather in opposition to those species, landscapes, and people narrativized as premodern.
The process of narrativizing modernity has always been a careful balancing act, concerned with projecting confidence and yet fraught with anxiety. Concern for the human place within deep time and what natural science might suggest about the ultimate fate of modern civilization (and, implicitly or sometimes explicitly, the white race) was a current that ran through natural historical discussion of modernity during the latter part of the nineteenth century. This tension has been on display throughout this dissertation as natural historians attempted to explain and demarcate the passage of time using animal species, human groups, and landscapes as temporal landmarks to orient themselves.History of Scienc
Swing and a Miss: Uncovering Behavioral Biases in MLB Betting Markets
No full textSports betting is a rapidly growing industry, gaining attention from sports fans, professional investors, and academia. Its attractiveness not only lies in the excitement about sports and potential rewards from betting itself, but also because it serves as a ground to study risk-taking and irrationality. This thesis investigates two of the most widely discussed biases, the favorite–longshot bias and momentum, within the MLB betting market. Using moneyline and run-line odds, as well as the evolution from opening to closing lines, we analyze whether these behavioral patterns persist in pricing and whether they offer exploitable trading opportunities. Our findings show that the favorite–longshot bias is clearly present in opening odds but disappears in closing odds, with stronger effects observed in home games. Momentum affects how teams are assigned to favorite or underdog categories, yet the market efficiently prices streak-related information. Similar patterns emerge in run-line markets. Although certain momentum-based signals appear promising for systematic betting, their economic value is too small to overcome transaction costs such as the bookmaker’s vig. Overall, this thesis provides a variety of evidence that the favorite-longshot and momentum bias are present in baseball betting markets, and that they stem from biases in bettor behaviors.Applied Mathematic
Functional and Structural Study of Alphavirus Receptors
No full textAlphaviruses are single-stranded, positive sense RNA viruses that can cause epidemic polyarthritis or encephalitis in humans and animals. Encephalitic alphaviruses include eastern equine encephalitis virus (EEEV), western equine encephalitis virus (WEEV), and Venezuelan equine encephalitis virus (VEEV). While EEEV and VEEV currently cause regionally contained outbreaks, their emergence into epidemics could be devastating. WEEV historically caused explosive epidemics of viral encephalitis in North America, but has since rapidly declined with the last case documented in the 1990s, a poorly understood process termed “submergence”. Alphaviruses enter host cells through attachment to specific protein receptors on the cell surface, a critical step targeted by antibodies and entry-blocking therapeutics. Here, we combine functional assays with in vitro, in vivo models and cryogenic electron microscopy (cryo-EM) to characterize the interaction of EEEV with its previously identified receptor, and of WEEV with newly identified receptors in this study. Our results provide insight into the ability of EEEV and WEEV to invade the central nervous system, and the molecular basis for the submergence of WEEV as a human and veterinary pathogen in North America, with implications for outbreak preparedness and the development of antiviral therapeutics.Virolog
Parallel Members – Parallelism, Translation, and Sacred Poetry, 1741- 1929
No full textParallel Members. Parallelism, Translation and Sacred Poetry, 1741-1929 is a comparative study of English and German translations and poetics of the Bible from the late 18th- to the early 20th-century. With chapters on Robert Lowth, Johann Gottfried Herder, Gerard Manley Hopkins and, finally, on Franz Rosenzweig and Martin Buber, its goal is to trace the pervasive influence of biblical parallelism on modern European poetics. In his 1741 Oxford lectures On the Sacred Poetry of the Hebrews, the English man of letters, Robert Lowth identified parallelismus membrorum as a chief structural principle of the language of the Old Testament: 'a certain equality, resemblance, or parallelism between the members of each period'. While ecclesiastic, rabbinic as well as academic sources commented on biblical parallelism more than once before Lowth’s work, it is thanks to their abundant reception that Lowth's lectures have often been viewed as the ‘invention’ of parallelism. Recent scholarly work on the influence of Lowth's concept has drawn on the resources of biblical criticism and the history of exegesis, on historiography as well as on rhetorical criticism, none has, however, investigated the implications of Lowth's parallelismus membrorum for modern European poetics and poetry. Neither biblical criticism, nor historiography, my dissertation reads specifically translations and poetics of the Bible to argue that parallelism has allowed for a radical reconceptualisation of poetic language in European modernity. Closely following the figures of parallelism at work in some of the readers of Lowth, who are themselves also translators of the Bible, my readings draw out the consequences of the skirmishes between parallelism, figuration and versification. I demonstrate that ever since its ‘invention’ in the middle of the 18th-century, the parallelism of Biblical poetry has served, again and again, as a way of reading and writing poetic language. The goal of my inquiry is to initiate a conversation on parallelism that will set novel, if necessarily parallel, ways for writing the literary history of modernity as well as for translation studies and Biblical poetics.Comparative Literatur
Studies on the Role of H3K36 Methylation using Histone Mutants
No full textChromatin is coated in a multitude of epigenetic marks of stereotyped distribution and unclear function. Among them, H3 lysine 36 trimethylation (H3K36me3) is enriched around actively transcribed gene bodies from yeast to humans, but its precise role remains disputed. Here, we interrogate the in vivo function of H3K36me3 using an innovative transgenic mouse model that expresses an inducible, dominant negative histone mutation leading to the genome-wide depletion of H3K36me3. Following induction of the mutation, mice became lethally anemic due to a profound differentiation block in terminal erythropoiesis. While H3K36me3 is typically associated with active chromatin, we found that loss of H3K36me3 does not reduce transcription levels, but rather increases transcription from the antisense strand of actively expressed genes. Importantly, these gains in antisense transcripts occurred specifically at genes with basal antisense transcription and robust levels of H3K36me3. Upregulated antisense transcripts triggered an interferon response that resulted in cell cycle arrest and blocked erythroid maturation. We further show that erythroblasts susceptible to elevated antisense transcription have markedly lower levels of DNA methylation and that other tissues are sensitive to antisense transcription when DNA methylation is inhibited. Based on these data, we propose that H3K36me3 and DNA methylation collaborate to repress antisense transcription at active gene bodies in mammals.Biological and Biomedical Science
Reptin is Required to Maintain Cardiomyocytes in a Proliferative Diploid State
No full textCardiomyocyte proliferation drives the addition of new myocardium during heart development and regeneration. We previously reported that the ATPase Reptin is a suppressor of cardiomyocyte proliferation in zebrafish as reptin mutant hearts show increased BrdU incorporation at 3 days post fertilization (dpf). In chapter 2, we determine that the ATPase function of Reptin is essential for its ability to repress cardiomyocyte proliferation at 3 dpf before making the paradigm shifting discovery that despite being hyperproliferative at 3 dpf, reptin-/- cardiomyocytes exhibit an anti-proliferative transcriptional profile. We then demonstrate that reptin-/- cardiomyocyte proliferation rates significantly decrease by 5 dpf and that reptin-/- cardiomyocytes accumulate in a polyploid state. The loss of Reptin drives cellular senescence but not an accumulation of DNA damage or apoptosis in cardiomyocytes. Through RNA-sequencing we identify several pathways and cellular processes which are dysregulated in reptin mutants including AP-1, tp53, cbx7a, and midbody/cytokinesis. We investigate the role of these mechanisms and their relationship to Reptin’s ability to regulate cardiomyocyte proliferation. We show that AP-1 signaling plays a role in controlling ventricular area in reptin mutants, but its effects on proliferation in reptin mutant hearts remain unclear. We further demonstrate that tp53 and cbx7a are not major drivers of the later stage cell cycle exit observed in reptin-/- cardiomyocytes. We further observe using human iPSC derived cardiomyocytes that Reptin localizes to the midbody of cytokinetic cardiomyocytes. We show that in mice, like zebrafish, Reptin is required for healthy heart development and the maintenance of cardiomyocyte proliferation. In the mouse, we show that the regulation of cardiomyocyte cell cycle exit by Reptin is cell autonomous and like in zebrafish loss of Reptin does not result in the significant induction of apoptosis. Our work has led to a new model whereby Reptin initially represses cardiomyocyte cell cycle entry in early heart development, but that it is required to maintain cardiomyocytes in a proliferative diploid state over the course of cardiac development. Since the major barrier to heart regeneration is cardiomyocyte cell cycle exit, therapeutic manipulation of Reptin could be used to stimulate renewal of the diseased heart.Biological and Biomedical Science
Mechanism of replisome-mediated parental histone inheritance
No full textInheritance of histones from parental DNA during DNA replication is important for the maintenance of epigenetic information and chromatin assembly. Decades of studies have demonstrated that parental histones are symmetrically segregated to the two newly synthesized daughter DNA strands. However, the mechanism(s) by which parental histones are inherited in coordination with DNA replication remain unclear. In my Ph.D. thesis, I use an inducible heterochromatin domain in Schizosaccharomyces pombe as a model system to study the requirement of a replisome-associated histone chaperone network for epigenetic maintenance of heterochromatin. Using this system, I demonstrated that several previously reported replisome-associated histone binding proteins and the histone chaperone FACT are required for epigenetic inheritance of heterochromatin. In addition, I discovered that epigenetic maintenance of heterochromatin requires the full fork protection complex consisting of Swi1, Swi3 and Mrc1 proteins. In particular, I showed that the non-essential DNA checkpoint factor Mrc1 is required for epigenetic inheritance through a newly identified conserved histone binding activity. Through collaborations examining the parental histone occupancy at newly synthesized DNA, we discovered that mutations in the Mrc1 histone-binding domain disrupt parental histone transfer to both the leading and lagging DNA strands. Using AlphaFold predictions, I found that the Mrc1 histone-binding domain preferentially locates at the center of the replication fork, suggesting that Mrc1 acts as a parental histone distribution site in the replisome.
With additional AlphaFold predictions, I identified multiple histone-binding domains and FACT-binding domains in the replisome. Several subcomplexes containing FACT and histones are predicted to form through these interaction interfaces. Genetic and biochemical experiments further confirmed that these newly identified interaction interfaces are required for heterochromatin maintenance. Collectively, these data led me to propose a model by which parental histones are captured by the FACT complex from the partially disassembled nucleosome induced by the replicative helicase and then inherited through multiple histone transfer intermediate sites at the replication fork.
My thesis work provides a mechanistic framework for studying the dynamics of parental histone inheritance at the replication fork. Future studies are needed to validate the presence of all parental histone transfer intermediate sites at the replication fork, determine the order of parental histone movements, and investigate the role of replisome-mediated parental histone inheritance in the development of multicellular organisms.Biological and Biomedical Science