Washington University Medical Center
Washington University St. Louis: Open ScholarshipNot a member yet
26344 research outputs found
Sort by
Probing the Mechanisms of Reinforcement Learning: Reinforcement Learning, Ventral Striatal Astrocytes, and the Dynamic Coordination of Information Seeking with Learning
No full textWhile reinforcement learning has been a vital component in artificial intelligence and machine learning, there exist many open questions about its implementations and how to improve them, in both minds and machines. Among these are i) the contribution of non-neuronal cell types to reinforcement learning, and ii) information-seeking behavior during reinforcement learning. In this thesis, we studied these main topics pertaining to reinforcement learning. In the first chapter, we examined the role of astrocytes in reinforcement learning, and in the second, we investigated human information seeking during reinforcement learning. Neurons in the human and animal brain have been known to support value-based decision-making and reinforcement learning in the brain [222, 131, 7]. However, to date, the role of astrocytes in reinforcement learning has not been thoroughly studied. Here, we trained mice on a bandit task, and attenuated the astrocyte activity in Ventral Striatum (VS), Dorsolateral Striatum (DLS), and Dorsomedial Striatum (DMS). We found that mice, whose astrocytes in VS were targeted, showed decreased bandit performance and increased win-stay behavior. Throughout computational modeling, we showed that these patterns could be explained by increased decision randomness. We then showed that these observations could be recapitulated by the deep neural network simulations, by attenuating input sharing across units. In the 2���� chapter, we investigate human information seeking during reinforcement learning. Information-seeking (IS) refers to intrinsic motivation towards obtaining information and is a critical aspect of human and animal behavior. While IS has been studied in the context of decision making, to the best of our knowledge, a thorough study of IS in the context of RL did not exist to date. We found that humans pay monetary rewards for obtaining information during RL. Moreover, their IS increases with reward uncertainty, decreases with learning, and correlates negatively with bandit performance. Furthermore, extensive reinforcement learning modeling revealed that IS correlates with the degree of random exploration, and having access to early information during learning increases the speed of learning in both humans and artificial neural networks. Overall, our work demonstrates a specific algorithmic relation between non-neuronal cell activity and reinforcement learning, and the coordination between human information seeking during reinforcement learning for the first time. xi
Epithelial Collective Response to Immune Cells, Matrix Proteins, and Mechanical Cues
No full textEpithelial cells form complex and dynamic tissue structures within the human body. During these processes, cells must come in contact with a variety of dissimilar cell types, extracellular matrix (ECM) proteins, and mechanical changes not aligned with homeostasis. In this thesis, we explore how changes to the surrounding environment can impact mechanoresponse of epithelial cells. We developed an in vitro system by using elastically tunable polyacrylamide substrates with collagen coating to understand how mechanics can affect collective cell migration. In the first aim, we added macrophages chemically polarized as M0, M1, or M2-like macrophages co-cultured with epithelial cells. We found that M0 and M1-like macrophages caused epithelial cells to cluster. These clusters were formed through an increase in contractility and a reduction of cell-ECM adhesions. When co-cultured with M2-like macrophages, epithelial cells maintained cohesive sheet-like structures and migrated along the substrate. This clustering phenotype arose because of macrophage-dependent cytokine secretion, competition between cell-cell and cell-ECM adhesions and increased epithelial cell contractility in the presence of M0 and M1-like macrophages. In the second aim, we sought to understand how microscale disruptions in heterogenous ECMs affect collective epithelial cell migration. Using a similar hydrogel set-up as the first aim, we used the elastically tunable polyacrylamide gels to coat either Collagen type-I or Collagen type-IV to the surface. When we generated microscale defects to the surface, cells on Collagen type-IV coatings collapsed into the superficial defect. The superficial defect caused a stalling in collective cell migration and multicellular collapse of the actin architecture. When we compared this to Collagen type-I coatings, we found that cells were able to migrate over the superficial microscale defect. Because Collagen type-I is more fibrillar compared to Collagen type-IV, cells were able to form longer filopodia to reach over the defect. Through both aims, we found mechanical tuning through cell stiffness, substrate stiffness, and membrane tension can change cellular response to immune cells and microscale defects. This work contributes to the growing understanding of the mechanisms that govern collective cell migration in response to microscale changes to the environment, both biomechanical and biochemical
Evaluation and Calibration of Coupled Low-Cost Particulate Matter Sensors
No full textLow-cost sensors (LCS), if used appropriately, are useful instruments to elucidate human exposure to particles suspended in the air (i.e. particulate matter, PM). Characterizing this exposure is crucial, as exposure to PM2.5 (i.e. particles with aerodynamic diameter ≤ 2.5 μm) negatively affects the respiratory, cardiovascular, and other organ systems and is the leading environmental burden for global mortality. Similarly, exposure to particles with aerodynamic diameters greater than 2.5 μm and less than or equal to 10 μm, here forth termed as PMCoarse, is known to cause health ailments for the upper respiratory system. This dissertation focuses on the MODULAIRTM-PM (MOD-PM), a device manufactured by QuantAQ that couples two light-scattering low-cost PM sensors–the Plantower PMS5003 and the Alphasense OPC-N3–to measure both PM2.5 and PMCoarse. QuantAQ purportedly uses the PMS5003 to estimate the mass concentration of particles below the detection limit of the OPC-N3; in this case, the MOD-PM uses both the PMS5003 and OPC-N3 to calculate PM2.5 estimates and uses only the OPC-N3 to calculate PMCoarse estimates. This brings to question how well the MOD-PM couples the various light-scattering measurements of these LCS to accurately estimate PM2.5 and PMCoarse mass concentrations. This dissertation presents research to evaluate and calibrate the coupling of the LCS used in the MOD-PM. The first chapter outlines the light-scattering principles underpinning the LCS used in the MOD-PM to examine the strengths and weaknesses of QuantAQ’s approach to coupling the PMS5003 and OPC-N3. A major limitation towards evaluating the PM estimates from the MOD-PM–the uncertainty of how the PMS5003 measurements are used to estimate the mass concentration of particles below the detection limit of the OPC-N3–is resolved by deriving an independent methodology to calibrate the PMS5003 to estimate PM mass concentrations for a size range of particles complementary to those detected by the OPC-N3. The second chapter utilizes this methodology in a locally weighted linear regression between PM estimates from MOD-PM and reference-grade devices to apportion error in PM2.5 and PM10 (i.e. PM2.5 + PMCoarse) estimates from the MOD-PM to the PMS5003 and OPC-N3. This regression demonstrates that the detection efficiency of the OPC-N3 causes this sensor to variably contribute to error in the PM2.5 and PM10 estimates of the MOD-PM. The third chapter utilizes gradient descent to calibrate the detection efficiency of the OPC-N3 by minimizing the error between MOD-PM and reference-grade estimates of PM2.5 and PMCoarse. This utilization of gradient descent lays the groundwork for future research attempting to infer calibration parameters of LCS through inverse modeling. Upon determining how to properly calibrate and couple the PMS5003 and OPC-N3, in the fourth chapter, these insights are used to measure the annual spatiotemporal trends of PM2.5 and PMCoarse from an outdoor network of fifteen MOD-PM devices in portions of the City of St. Louis and nearby north St. Louis County. This pragmatic use of the MOD-PM requires the methodologies derived to calibrate the PMS5003 and OPC-N3 to be repurposed to address practical issues affecting MOD-PM measurement error such as seasonal variations and sensor degradation. In addition to this research on coupled low-cost PM sensors, the fifth chapter describes work to improve and demonstrate the Multichannel Organics In situ enviRonmental Analyzer (MOIRA) for mobile platform measurements of volatile organic compounds
The influence of host genetics and microbial interactions on the gut microbiome composition of a wild baboon population
No full textUnderstanding how evolutionary processes modulate the complex interplay between host genetics, behavior, and the gastrointestinal (GI) microbiome is fundamental to ecological and evolutionary research. This dissertation explores the drivers of multi-kingdom GI microbiome and virome structure and dynamics within a wild baboon hybrid zone, emphasizing the co-evolutionary dynamics both within microbial communities and between hosts and their associated microbes. Chapter One reviews the literature, establishing the critical role of hybridization in shaping primate immune systems and introducing Kinda-Grayfooted chacma hybrid baboons as a compelling model for studying host-microbe co-evolution amid genomic admixture. This chapter details how immune gene introgression, a common consequence of hybridization, can influence host physiology relevant to the GI ecosystem. Furthermore, it examines the composition and function of the gut microbiome, highlighting how microbial diversity and interactions are affected by hybridization and immune gene variations. By integrating these perspectives, the chapter underscores the complex relationships between host genetics, immune responses, and microbial communities within the GI tract. Chapter Two examines the relative contributions of host genetic variation (derived from hybridization), environmental factors, and social behavior, including sex, in shaping overall GI microbiome composition and diversity. I found that direct host genetic ancestry, represented by hybrid category, had a limited impact on overall microbiome structure. Instead, temporal environmental shifts emerged as predominant ecological filters, driving significant changes in microbial diversity and composition. Additionally, biased social interactions, such as mating and grooming, along with sex differences, played pivotal roles in microbial transmission. These interactions contributed to similarities in GI microbiota among frequently interacting individuals, thereby shaping a social microbiome. Chapter Three delves into the nuanced intra-kingdom microbial relationships, emphasizing niche construction as a central organizing principle. This chapter reveals complex intra-specific cooperation (e.g., quorum sensing, biofilm formation) and competition, alongside diverse inter-specific mutualistic and antagonistic relationships operating within the bacterial, eukaryotic, and viral communities respectively. It highlights how these internal dynamics contribute to the stability and functional capabilities of each microbial kingdom. Chapter Four extends this analysis to inter-kingdom microbial relationships, presenting multi-microbial network analyses that revealed distinct architectures. Bacteria-eukaryot
Spectral Communities: Spanish-Language Practices in Nineteenth-Century St. Louis
No full textThis dissertation examines the Spanish-language practices that developed in the context of U.S. economic expansionism toward Latin America from the last quarter of the nineteenth century to the beginning of the twentieth century. The study centers on St. Louis, Missouri, a driving force behind this commercial crusade, and pieces together the city’s rich Spanish-language archive, comprising periodical publications, educational publishing, Latin American clubs, translation, and language teaching. I suggest that within this myriad of practices lie the untold stories of belonging, negotiation, and veiled resistance of the Spanish speakers who participated in this hemispheric drama as teachers, interpreters, and translators. Employing archival research, microhistory, and the cultural analysis of these practices, I argue that Spanish speakers strategically used their linguistic competence to find a space of enfranchisement in the U.S. American national project. Furthermore, I maintain that the public performance of Spanish by the Anglo-American participants in these clubs served as an important laboratory for hemispheric and imperial imaginings; a practice for grappling with the heightened connectivity between the U.S. and the Western Hemisphere towards the end of the nineteenth century. This project analyzes Spanish as practiced by both L1 and L2 speakers in the nineteenth century and thus provides insight into the co-constructedness of Spanish-language ideologies in the U.S. during that period
The Case for Automatic Enrollment in Trump Accounts: Lessons from Maine’s CDA
No full textIn 2014, Maine\u27s statewide Child Development Policy, the My Alfond Grant, transitioned from an opt-in enrollment structure to an automatic one. Written as federal policymakers weigh enrollment options for the new Trump Accounts, this brief highlights lessons from Maine\u27s transition and findings from the the SEED for Oklahoma Kids experiment, offering recommendations for implementing the new federal policy
Intravital Optical Imaging of Peripheral Nerve Injury Response
No full textPeripheral nerve injuries in animal models require imaging approaches that can reveal the dynamic, multiscale process of regeneration over time without disrupting repair mechanisms. Traditional methods lack the ability to visualize cellular behavior and matrix remodeling continuously. We present an imaging technique based on an implantable nerve window and fluorescent labeling, enabling high-resolution, longitudinal cellular-scale observation of peripheral nerve regeneration in living mice. We first evaluated near-infrared fluorophore-conjugated fibrin as a transient marker for early regenerative tissue within nerve guidance conduits. Imaging performed upon surgical re-exposure of the nerve revealed the incorporation of fibrin into early-stage regenerative cables, with gradual signal degradation over two weeks. This approach demonstrated the feasibility of using extracellular matrix-based tracers for dynamic visualization of regenerative tissue and motivated the development of longer-term imaging strategies. To enable stable, chronic optical access to the sciatic nerve in a mouse model, we developed a surgical method to expose the nerve and implant a permanent, polydimethylsiloxane (PDMS) skin-embedded window, which outperformed rigid, 3D-printed designs. This technique allowed imaging of the nerve for more than 90 days post-implantation, with no detectable structural or functional impact, minimal inflammation, and limited fibrotic response, permitting continuous observation of regeneration within the same animal. Using this platform, we achieved multiplexed, in vivo imaging of the peripheral nerve environment by combining a transgenic fluorescent reporter (Thy1-YFP for axons, S100-GFP for Schwann cells, Tie2-GFP for vascular endothelial cells) with optional additional labeling (second-harmonic generation for collagen, Nile Red for myelin/lipids, and/or fluorophore-conjugated tomato lectin for vasculature). We applied the nerve window across multiple injury models, including compression, crush, partial transection, and gap transection followed by conduit repair, to longitudinally observe distinct regenerative responses, such as immune infiltration, axonal degeneration, Schwann cell migration, axon–collagen interactions, and vascular changes. We further refined the nerve window approach in the conduit repair model by testing conduit technologies and tracking regeneration in different mouse strains. FVB/N mice exhibited less fibrosis but also reduced regenerative capacity compared to C57BL/6 and B6D2. Statistical evaluation showed that early-stage visual observations could predict later regenerative progress. Customized PDMS and poly(glycerol sebacate) conduits were successfully imaged and supported gap regeneration. Finally, we tested fluorescent fibrin as a passive reporter scaffold within the nerve repair conduit lumen and showed that it exhibited gradual degradation in vivo. Together, this work introduces a system combining in vivo optical imaging, multimodal fluorescent labeling, and adaptable injury models to enable noninvasive, high-resolution tracking of peripheral nerve regeneration in live mice. These techniques open the door to mechanistic investigation of regenerative processes at a cellular scale and accelerated preclinical evaluation of therapies targeting peripheral nerve repair
Large-Scale Modeling and Inference of Non-Stationary Brain Dynamics
No full textNeural modeling has long been a tool for understanding the brain\u27s function, in both healthy subjects and those experiencing neural pathologies. These models have captured structures in the brain ranging in spatial scale from the subcellular level to the whole-brain level, and processes ranging in temporal scale from milliseconds to multiple hours. Currently, however, there are very few dynamical systems models of whole-brain activity, and the modeling methodologies that do exist rely on neural data modalities which are impractical for use in clinical settings. At the same time, a wealth of data is being generated by patients undergoing clinical monitoring. Therefore, there is an opportunity for engineering modeling techniques that make use of this data and capture the latent brain dynamics of the person being monitored. A key challenge in creating these modeling techniques is the fact that neural data is nonstationary, varying with time and physiologic state. Indeed, the current state-of-the-art techniques for dynamical systems modeling of the brain are formulated for stationary settings, and cannot accurately capture dynamics which change over time. In this dissertation, I develop data-driven methods for creating individualized models of neural activity which can vary temporally according to neurophysiologic states latent to the neural dynamics. I begin by adapting a current method for modeling whole-brain dynamics for use with electroencephalography (EEG) data, in order to gain mechanistic insights into changes in brain dynamics during events of clinical significance. I then extend this method to capture changing dynamic regimes when the switches in dynamic regime are labeled. This is accomplished by modeling, in essence, the process of neuromodulation. Here, instead of a static connectivity matrix typical of neural network models, I deploy a constant base connectivity matrix multiplied by a switched modulation matrix. Finally, I combine this modeling methodology with a method for blind identification of switches in neural state, in order to describe and infer the changes in dynamic regime present in non-stationary neural data, as well as the dynamics of each individual regime. The methods presented in this dissertation represent an important step forward in modeling and understanding non-stationary brain dynamics on an individual level
Understanding the Role of Hope on the African American Identity through James VanDerZee\u27s Wedding Day Harlem, 1926
No full textThis essay explores how hope among migrants of the Great Migration was shaped by positive depictions of African Americans in photography, particularly through the work of Harlem Renaissance photographer James VanDerZee. It argues that these images offered migrants a vision of their desired futures, but true hope depended on the actions taken to realize those visions
Breaking the Ice: Disability Representation in the “Ice” Love, Death + Robots Adaptation
No full textThis essay examines how Netflix’s Love, Death + Robots (LDR) adaptation of Rich Larson’s short story “Ice” transforms its original sibling dynamic into a commentary steeped in ableist tropes. While Larson’s story portrays Sedgewick, the unmodded brother, as autonomous and confident, the episode reinterprets him through infantilizing and victimizing lenses, casting him as a disabled figure in need of rescue. The LDR episode’s decisions to downplay or embellish certain elements of Larson’s short story reveal eco-ableist and transhumanist-ableist themes that link bodily limitation to environmental incompatibility and valorize technological modification as physical superiority. Furthermore, Fletcher’s deceptive “help” exemplifies the heroic supporter trope, reframing paternalistic intervention as empowerment. By contrasting the short story’s nuanced portrayal of agency with the episode’s reliance on supercrip and heroic helper narratives, this essay reveals the need for nuanced disability representation that subverts common disability tropes rather than reinforcing them