71446 research outputs found
Sort by
Error-corrected quantum processing with neutral atoms
Get PDFQuantum computers open new scientific avenues, from exploring complex quantum mechanical systems to new computational paradigms, but face the fundamental challenge of decoherence. Remarkably, decoherence can be prevented by creating highly entangled states of physical qubits that encode an error-corrected “logical” qubit. This thesis will describe the development of quantum computing with reconfigurable arrays of neutral atoms and their use for quantum processing with logical qubits. Quantum processing in this approach is based on the coherent transport of atoms shuttled by optical tweezers, enabling any-to-any connectivity, high-fidelity programmable logic, and mid-circuit processing within a zoned architecture. Logical qubit processing is greatly facilitated by parallel control and transversal operations, and is used for experiments ranging from entangling logical qubits to their use for precise simulation of quantum scrambling. Core physical mechanisms for achieving deep-circuit, universal algorithms with logical qubits are identified, and these are leveraged into new techniques that greatly reduce overheads for large-scale computation. Finally, applications of quantum processors will be explored via analog and gate-based quantum simulations. These results, alongside other recent advances, herald a transition to error-corrected quantum processing, establishing foundations that can enable future large-scale quantum computers and their useful applications.Physic
Facing the Monster: The Ethical Challenge of Weird Fiction
Get PDFThis dissertation argues that the prioritization of empathy as the foundation of ethical reading, especially through identification, obscures the alterity of the other and forecloses truly ethical engagement. Through close readings of texts that center the literary monster as a site of estrangement, I propose that Weird fiction, defined not as a genre but as a mode, disrupts empathic strategies and creates affordances for ethical reading rooted in attentiveness, humility, and refusal of mastery. Chapter One establishes a theoretical framework grounded in Emmanuel Levinas’s ethics of the face and James J. Gibson’s theory of affordances to critique aesthetic mechanisms of empathy. Chapter Two examines the fiction of Mariana Enriquez, whose grotesque figures resist empathic access but demand ethical attention despite discomfort. Chapter Three turns to Antoine Volodine’s post-exoticism, in which formal and meta-fictional hostility disorients the reader and renders them complicit as a spectator of violence as well as intruder. Chapter Four explores Old French texts by Marie de France and Chrétien de Troyes where shapeshifters, werewolves, and other monsters prefigure Weird encounters that unsettle chivalric epistemologies and challenge the legibility of the human. The conclusion reflects on the pedagogical and political implications of ethical estrangement, proposing a critical posture for reading and teaching literature that centers the irreducible other. Across these chapters, I argue that the Weird monster functions as an interruption; it is an aesthetic, ethical, and affective provocation that challenges reading for identification. I advocate for an ethics that, rather than beginning with empathy and falling victim to its limitations, allows one to turn their attention towards the face of the monster.Romance Languages and Literature
A Causal Inference Framework for Identifying Critical Windows of Time-Varying Exposures
Get PDFThere has been great clinical interest in the concept of ‘critical windows’ or ‘sensitive periods’ of environmental exposures. The concept of a critical window, defined formally in this dissertation, refers to a specific time period during which an individual is more susceptible to developing an outcome in response to a particular exposure than at other times. The statistical methods used to identify these critical windows have several limitations, and applied researchers are often limited to using methods developed for different research questions which can lead to bias, inflated Type I error rates, and low power.
Studies on environmental exposures are almost always observational by necessity, and it remains an ongoing challenge to interpret results as the causal effect of intervening on the exposure rather than merely as an association between the exposure and the outcome. Because the methods currently in use have not been previously examined through a causal inference lens, results across studies are difficult to compare, even if the same covariates are used.
This dissertation seeks to combine these two areas of interest by proposing a framework for the identification of critical windows from a causal inference perspective. Throughout this work, we demonstrate how different methods should be employed to answer subtly different research questions and compare our methods to existing approaches through simulations where appropriate.
In Chapter 1, we introduce our novel flexible CAusaL Identification of Critical windOws - Modified Treatment Policy (CALICO-MTP) framework, extending previous work on using a dose modification scheme to estimate the causal effect of continuous exposures. We propose dividing the concept of critical window identification into three distinct research questions, each addressed with different approaches. These questions are: 1) Curve estimation: what does the exposure-outcome relationship look like over time? 2) Hypothesis testing: is there any time window during which there is an effect of intervening on the exposure? and 3) Window selection: after determining that there is a causal relationship, what is the critical window for that exposure? For the first question, we propose a curve estimation strategy to yield results similar to those of the commonly used distributed lag model (DLM). For the second, we propose estimating the effect of intervening on
all biologically plausible windows and combining the p-values using the Aggregated Cauchy Association Test (ACAT), a p-value combination method that accounts for strong correlations between test statistics. For the third, we discuss strategies for selecting the window once the global null has been rejected. We apply these methods to a dataset from Beth Israel Deaconess Medical Center (BIDMC) and compare them to previous results regarding the effect of Nitrogen Dioxide (NO2) exposure on the 32-40 week fetal head circumference as measured by ultrasound45, and we present a novel visualization for the causal effect of intervening on time intervals.
In Chapter 2, we present a variant of this framework, CALICO-ADRF, that explicitly models nonlinear dose-response relationships by estimating the Average Dose Response Function (ADRF) for each time window. This nonlinear relationship is particularly relevant for environmental exposures such as metals, where some are necessary minerals at low exposures but act as toxins at high exposure levels, and temperature, which may exhibit a thresholding effect for certain outcomes. We use a scalar test statistic that is the integrated squared derivative of the estimated ADRF to perform global hypothesis testing with Type I error control and improved power compared to the methods of Chapter 1 for biologically-plausible nonlinear dose-response curves. We demonstrate these results looking at the effect of maternal prenatal temperature exposure and birthweight for full-term deliveries in the same BIDMC cohort.
In Chapter 3, we present a discussion of causal inference concepts specifically tailored to the methods most commonly used for time-varying environmental exposures, offering a novel perspec- tive for researchers. We present a framework through which the target estimand of different mod- eling approaches can be compared, improving the ability to draw meaningful and comparable conclusions across studies. We explore the different estimands that these methods target and illustrate when these estimands align or diverge depending on the underlying causal structure of the exposure. Finally, we provide guidance for researchers on how to appropriately align their methodological choices with their research questions.Biostatistic
Statistical Methods for Negative-Unlabeled Data with Application to Long COVID
Get PDFNegative-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
Leveraging PhagoID to Define How Cytokines Reprogram the Phagosomal Proteome
Get PDFThe ability of phagocytosing diverse cargoes and maintaining tissue homeostasis under different immune contexts and challenges is what lends tissue resident macrophages their identity as immune sentinels. Previous studies that were aimed towards isolating and studying the macrophage phagosome suffered from certain limitations due to the highly dynamic and interactive nature of this organelle, making its proteomic analysis an especially arduous task. In a recent study, we reported the development of a novel tool called PhagoID which uses a proximity labeling-based strategy for resolving phagosomal lumen proteins with high specificity and efficiency, facilitating their identification using mass spectrometry and downstream proteomic analyses. Here, we validate the use of epigallocatechin gallate (EGCG) as a ROS scavenger capable of minimizing extracellular labeling in the case of PhagoID. Next, we adapted PhagoID to quantify the phagosome lumen proteome of fetal-derived alveolar-like macrophages (FLAMs) under the effect of type I interferon signaling.
The phagocytic activity of alveolar macrophages (AMs) is important both at baseline for routine cleaning of cellular debris, as well as to initiate a strong inflammatory responses to respiratory pathogens. We profiled the phagosomal proteome of FLAMs as a model for primary AMs and found proteins involved in the class I MHC pathway to be the only ER subset enriched in the phagosome on IFN-β activated FLAMs. Additionally, we identified the presence of proteins belonging to 3 families of interferon induced GTPases (IIGPs), including multiple GBPs and all IRGM proteins which have been well recognized for their role in cell autonomous immunity against bacterial infections like Mycobacterium tuberculosis and Chlamydia trachomatis. We also detected the enrichment of the enzyme aconitate decarboxylase (ACOD1/IRG1) involved in the production of the anti-microbial metabolite, itaconate. Going ahead, we propose the design of an itaconate biosensor targeted to the endosomal and lysosomal compartments to understand which host factors govern itaconate trafficking to phagosomes. Thus, using PhagoID enabled us to gain important insights into how IFN-β, a cytokine known for its highly context-dependent role in case of bacterial infections, can alter the phagosome proteome of AMs.Medical Scienc
Internal state modulation of striatal dopamine signaling
No full textDopamine (DA) concentration in the striatum fluctuates on two timescales: fast, sub-second
“phasic” changes, and slow, minutes to hours long, shifts in the “tonic” baseline. Phasic striatal DA
fluctuations may represent a prediction error signal, or the difference between what an agent expects
to happen and what actually happens. Animals are thought to implement this prediction error in a
temporal difference learning framework to update policies mapping states to actions to learn how
to attain rewards and avoid threats in the environment. Although much progress has been made
to understand the heterogeneity of this phasic signal across striatal subregions and how different
stimuli evoke different phasic dopaminergic signals, much less is known on the role that internal
state of the agent plays in shaping phasic and tonic DA signaling.
To investigate how internal state modulates phasic and tonic DA signaling, I employed fluorescence
lifetime photometry at high temporal resolution (FLiP-R) coupled with novel DA sensors to measure
absolute levels of DA in the striatum. I conducted recordings and manipulations of DA signaling
in two striatal subregions - the nucleus accumbens core (NAC) in which phasic DA signaling
is thought to represent a reward-prediction error, and the tail of striatum (TS) in which phasic DA
signaling is thought to represent a threat-prediction error. In the TS, hunger increases tonic DA
while suppressing the phasic TS DA response to modulate exploration of novel, potentially threatening
stimuli. The hunger signal that modulates the phasic TS DA signaling pathway derives from
the activity of hypothalamic agouti-related peptide (AgRP) neurons. In the NAC, hunger increases
the tonic DA level, which in turn modulates an animal’s motivation to work for a fixed reward. I
thus delineate how phasic and tonic DA signaling integrates internal state across different striatal
subregions to modulate different aspects of behavior.Neuroscienc
Emperor and Physician: Plague, Politics, and Healing in the Justinianic Novels
Get PDFThe first bubonic plague outbreak, the “Plague of Justinian,” has been subject to scrutiny to determine how it impacted the cultural, political, and economic trajectory of the Byzantine Empire. Central to this debate is the question of the plague’s minimal mention in non-literary sources, particularly the “Novels,” a series of laws released by Emperor Justinian throughout his reign. In recent years, the emperor’s relative silence on the matter has been cited as evidence that the plague was not significant enough to merit a robust imperial response. While the Novels make virtually no explicit reference to plague, this paper identifies a shift in legal language during the early years of the pandemic wherein the emperor offered a series of allusions likening himself to a physician practicing his craft. These allusions are forceful in their comparative quality, declaring an equivalency of lawmaking and healing. It is the position of this paper that the emperor used his legislation as a vehicle to modify his public image, embodying the mindset of a physician to underline his virtuous philanthropy and offset criticism throughout the plague. Compared especially to his successors, Justinian’s language was uniquely deployed and represents an understudied political effect of the pandemic.Extension Studie
Large Language Model Embeddings for Single-Cell Transcriptomics: A Framework for Robust Classification of Motor Neuron Vulnerability
No full textSelective vulnerability is a defining feature of neurodegenerative disease, yet
conventional single-cell analysis often blur subtype boundaries and obscure the molecular
programs that distinguish resilience from decline. To address these limitations, this study
introduces an integrative framework that combines large language model (LLM)-derived gene
embeddings with contrastive learning to generate biologically contextualized representations of
single-cell RNA sequencing data.
The framework was applied to retinal ganglion cells (RGCs) and validated in motor
neurons (MNs) to assess generalizability across neuronal contexts. For each cell, the most highly
expressed genes were linked to curated textual summaries and embedded using pretrained
transformer models. Expression-weighted aggregation produced cell-level representations, while
contrastive learning further refined these embeddings by isolating subtype-specific
transcriptional features from background expression.
Across embedding architectures, LLM-based representations consistently outperformed
graph-based baselines in MN subtype classification and preserved rare RGC subtypes under
data-limited conditions. The same embedding and contrastive framework transferred reliably
across datasets through KNN-based mapping, reproducing subtype topology without additional
training. Visualization and cross-referencing analyses revealed transcriptional gradients
consistent with known vulnerability hierarchies and highlighted pathways associated with
neuronal differentiation, metabolic resilience, and degeneration.
Collectively, these findings demonstrate that integrating literature-derived biological
context with self-supervised learning provides a scalable framework for investigating the
molecular logic of selective neuronal vulnerability in neurodegenerative disease.Computer Scienc
Oncogenic and epigenetic inhibitors kill castration-resistant prostate cancer by cooperatively suppressing energy metabolism
No full textProstate cancer is the second leading cause of cancer-related deaths in men. Most patients diagnosed with advanced prostate cancer develop resistance to first-line treatments, and eventually develop incurable castration-resistant prostate cancer (CRPC). Thus, there is an unmet clinical need to develop more effective therapies for this disease. This dissertation presents co-targeting the oncogenic PI3K/AKT signaling pathway and the epigenetic regulator EZH2 as an effective combinatorial strategy to kill castration-resistant prostate cancer in pre-clinical models. The PI3K signaling pathway in overactive in 70% of CRPC tumors, and defects in the pathway are linked to disease progression. EZH2 is also overexpressed in metastatic prostate cancer, correlates with lower failure-free survival, and has been shown to induce metastasis in mouse models of prostate cancer. In my dissertation, we demonstrate that EZH2 inhibitor synergizes with multiple PI3K signaling pathway inhibitors to kill CRPC in vitro, and causes tumor regression in vivo. Unbiased transcriptional and metabolic approaches revealed that EZH2 and PI3K/AKT signaling inhibitors kill CRPC by cooperatively suppressing glycolysis and oxidative phosphorylation, causing devastating energy crisis. Additional mechanistic and functional studies demonstrated that PI3K pathway and EZH2 inhibitors cooperatively suppress these metabolic pathways by dramatically decreasing the critical metabolic regulators HIF-1A and MYC at the protein level. In parallel, EZH2 inhibition upregulates the pro-apoptotic sensor protein BMF, which triggers apoptosis in response to the metabolic stress. Additionally, I also discuss some preliminary positive and negative results aimed at elucidating the mechanism by which EZH2 inhibitors suppress MYC expression and metabolism. Specifically, I propose that EZH2 inhibition leads to a feedforward loop between decreased MYC translation and amino acid transport. Moreover, I report a list of direct EZH2 targets in CRPC, which includes FOXO1 and FOXO3 as potential critical regulators of cell death in response to EZH2 and PI3K pathway inhibitors. Together, the data presented in this dissertation reveals a promising therapeutic strategy for CRPC, and demonstrates how energy metabolism can be fatally suppressed by targeting upstream oncogenic and epigenetic nodes.Biological and Biomedical Science
Real-World Evidence in Oncology Using Reference Trial Emulation Across Multiple Electronic Health Record Databases
No full textBackground: Oncology specialty electronic health record (EHR) databases are increasingly used to generate real-world evidence (RWE) in prognostic modeling and comparative effectiveness research. These data sources offer the potential to complement randomized clinical trials (RCTs) by reflecting routine clinical practice and including more heterogeneous patient populations. However, their use remains challenged by incomplete capture of key prognostic variables, missing data, and heterogeneity across databases. As a result, it remains uncertain under what conditions oncology EHR data are sufficiently complete and reliable to support valid prognostic assessment and causal inference. Benchmarking real-world analyses against RCTs has been proposed as an approach for assessing fitness-for-purpose of a given real-world data source for specific, closely related comparative effectiveness questions.
Objectives: The objectives of this body of work were to: (1) evaluate the generalizability and performance of a computable prognostic score for overall survival (OS) across multiple oncology specialty EHR-derived databases; and (2)/(3) explore the extent to which multiple oncology specialty EHR-derived data can replicate OS treatment effects observed in two oncology RCTs.
Methods: First, we evaluated the performance of the Real-wOrld PROgnostic (ROPRO) score, a multivariable prognostic model derived from EHR-derived data, across four independent oncology EHR databases covering multiple cancer types and disease settings. ROPRO scores were computed using routinely collected demographic, clinical, and laboratory variables, with missing covariates imputed using random forest-based methods. OS was modeled using Cox proportional hazards models, and performance was assessed using discrimination and calibration metrics.
Second, we conducted two exploratory comparative effectiveness studies emulating the MONARCH-3 and MONALEESA-2 RCTs, which evaluated cyclin-dependent kinase (CDK) 4/6 inhibitors combined with endocrine therapy as first-line treatment for hormone receptor-positive, human epidermal growth factor receptor 2 (HER2)-negative metastatic breast cancer. Across three oncology specialty EHR-derived databases, we aligned eligibility criteria, treatment definitions, and follow-up with the reference trials. OS was the outcome of interest. Missing data were addressed using multiple imputation by chained equations, and confounding was controlled using 1:1 nearest-neighbor propensity-score (PS) matching. Databases achieving covariate balance after PS matching were included in the primary analysis, which included a fixed-effects inverse-variance meta-analytic approach. All analyses were conducted under preregistered protocols and were considered exploratory due to limited statistical power.
Results: The ROPRO demonstrated consistent and generalizable prognostic performance across databases and cancer types when all or nearly all variables were available, with moderate-to-good discrimination and overall adequate calibration. Performance deteriorated in a database with limited variable availability, highlighting the sensitivity of prognostic modeling to data completeness.
In the MONARCH-3 emulation, two databases met inclusion criteria for the primary analysis, and the pooled OS hazard ratio (HR) was closely aligned with the randomized trial estimate, although database-specific estimates varied substantially in magnitude and direction. In the MONALEESA-2 emulation, only one database achieved covariate balance; the resulting effect estimate was aligned with the trial result, while additional databases were informative only in sensitivity analyses. Across both emulations, pooled estimates generally approximated trial findings, but uncertainty was substantial and individual databases produced heterogeneous results.
Conclusions: Collectively, these studies demonstrate that oncology EHR-derived real-world data can, under certain conditions, support valid prediction and approximate randomized trial estimates for OS. However, fitness-for-purpose is highly context specific and depends critically on data completeness, mortality capture, confounder availability, and line-of-therapy curation. The ROPRO score was generalizable across datasets when key variables are well captured, while confidence in real-world comparative effectiveness analyses may be increased through benchmarking against closely related RCTs. These findings underscore the need for pre-specified study design and analysis, transparent reporting, and database- and question-specific diagnostics to support credible prediction and causal inference in oncology RWE.
: First, we evaluated the performance of the Real-wOrld PROgnostic (ROPRO) score, a multivariable prognostic model derived from HER-derived data, across four independent oncology EHR databases covering multiple cancer types and disease settings. ROPRO scores were computed using routinely collected demographic, clinical, and laboratory variables, with missing covariates imputed using random forest-based methods. OS was modeled using Cox proportional hazards models, and performance was assessed using discrimination and calibration metrics.
Second, we conducted two exploratory comparative effectiveness studies emulating the MONARCH-3 and MONALEESA-2 RCTs, which evaluated CDK4/6 inhibitors combined with endocrine therapy as first-line treatment for hormone receptor-positive, human epidermal growth factor receptor 2 (HER2)-negative metastatic breast cancer. Across three oncology specialty EHR-derived databases, we aligned eligibility criteria, treatment definitions, and follow-up with the reference trials. OS was the outcome of interest. Missing data were addressed using multiple imputation by chained equations, and confounding was controlled using 1:1 nearest-neighbor propensity-score (PS) matching. Databases achieving covariate balance after PS matching were included in the primary analysis, which included a fixed-effects inverse-variance meta-analytic approach. All analyses were conducted under preregistered protocols and were considered exploratory due to limited statistical power.
Results: The ROPRO demonstrated consistent and generalizable prognostic performance across databases and cancer types when all or near-all variables were available, with moderate-to-good discrimination and overall adequate calibration. Performance deteriorated in a database with limited variable availability, highlighting the sensitivity of prognostic modeling to data completeness.
In the MONARCH-3 emulation, two databases met inclusion criteria for the primary analysis, and the pooled OS hazard ratio (HR) was closely aligned with the randomized trial estimate, although database-specific estimates varied substantially in magnitude and direction. In the MONALEESA-2 emulation, only one database achieved covariate balance; the resulting effect estimate was aligned with the trial result, while additional databases were informative only in sensitivity analyses. Across both emulations, pooled estimates generally approximated trial findings, but uncertainty was substantial and individual databases produced heterogeneous results.
Conclusions: Collectively, these studies demonstrate that oncology EHR-derived real-world data can, under certain conditions, support valid prognostic modeling and approximate randomized trial estimates for OS. However, fitness-for-purpose is highly context specific and depends critically on data completeness, mortality capture, confounder availability, and line-of-therapy curation. The ROPRO are generalizable across datasets when key variables are well captured, while comparative effectiveness analyses may benefit from routine benchmarking against closely related RCTs. These findings underscore the need for pre-specified study design and analysis, transparent reporting, and database- and question-specific diagnostics to support credible causal inference and prognostic assessment in oncology RWE.Population Health Science