UC San Diego

We investigated phytoplankton growth and micro- and mesozooplankton grazing patterns in the California Current Ecosystem (CCE) during summer 2021. Two water parcels, followed over a duration of 4-5 days using satellite-tracked drogued drifter for quasi- Lagrangian experimental cycles were investigated for inshore and offshore differences. Phytoplankton growth rates and microzooplankton grazing rates were determined using the two-point dilution method, and daily Bongo tows were deployed for mesozooplankton collection, for biomass and grazing estimates based on gut fluorescence. Instantaneous rates of growth and grazing between the two cycles were µ = 0.45 (± 0.13) d-1 for Cycle 2 (inshore) and 0.60 (± 0.1) d-1 for Cycle 3 (offshore), and microzooplankton grazing rates were 0.36 (± 0.21) d-1 for Cycle 2 and 0.37 (± 0.11) d-1 for Cycle 3. Mesozooplankton contributed much less to grazing for both cycles, grazing 0.05 (± 0.02) d-1 inshore and 0.025 (± 0.006) d-1 offshore, removing 4% and 2% of phytoplankton standing stock, respectively. In both cycles, the net calculated phytoplankton growth was positive, but this was only statistically significant for the offshore region. The dominant grazers within the mesozooplankton community were not consistent between the two regions of the CCE: the three smallest size classes (0.2-2 mm) contributed the most to grazing in Cycle 2, while in Cycle 3, the dominant grazers were the two smallest size classes (0.2-1 mm). Mesozooplankton grazing showed changes related to diel vertical migration. These analyses contribute to our understanding of growth and grazing dynamics in plankton food webs, and to understanding carbon cycling in the CCE.

This dissertation consists of four chapters, each of which studies monetary policy, financial stability, or their interaction.

Chapter one shows empirically that, contrary to theoretical claims, raising interest rates increases bank leverage. I propose and empirically validate the loan-loss mechanism to explain this result: contractionary shocks increase loan losses, reduce bank profits and equity, and ultimately increase bank leverage. I develop a banking model where floating-rate loans entail a trade-off between interest rate risk and credit risk, which generates the loan-loss mechanism. Using microdata, I provide empirical evidence consistent with floating-rate loans hedging interest rate risk at the expense of generating loan losses.

Chapter two examines the effects of central bank meetings on stock returns. Cieslak et al. (2019) show that stock returns in the US and internationally are driven by even-week meetings of the Federal Open Market Committee. I find that the US result and the proposed mechanism do not hold out-of-sample, losing robustness as early as 2004. Prior to 2004, there appear to be outliers driving the result. Finally, I show that the international result does not apply in either the UK or Japan.

Chapter three studies the consequences of including financial stability among the central bank's objectives when market players are strategic. Our model predicts that central banks underreact to economic shocks, a prediction consistent with the Federal Reserve’s behaviour during the 2023 banking crisis. Policymakers’ stability concerns bias investors' choices, inducing inefficiency. If central banks have private information about their policy intentions, the equilibrium forward guidance is vague because fully informative communication is not credible. A "kitish" central banker, who is less concerned about stability, reduces these inefficiencies.

Chapter four studies how financial and production networks affect the transmission of financial shocks to the real economy. We propose a general equilibrium model of production networks featuring heterogeneous banks and endogenous firm-bank linkages. We theoretically characterise the aggregate effects of bank-specific shocks in terms of a number of sufficient statistics. We suggest an approach to empirically complement our theoretical framework which relies on misconduct provisions of UK banks combined with detailed firm-bank-loan data to construct instruments for firms’ credit supply.

Surface waves are a subset of seismic waves that travel along the surface of Earth. These waves can encounter different subsurface structures depending on their azimuth, which can alter their velocity. This directional dependence of velocity on azimuth is called azimuthal anisotropy. We examine this property along with the velocity itself to gather insight into the structure of the lithosphere and the deformation of the mantle. We used Rayleigh waveform data from the USArray Transportable Array (TA), a ~70 km grid spaced seismic network that migrated across the contiguous United States from 2004-2015. Using these stations, we applied a sub-array technique that consists of a target station and its four nearest neighboring stations. We then determined the average frequency-dependent phase velocity within the sub-array for each individual earthquake from a large list of suitable earthquakes to ensure quality of data and gain understanding of the azimuthal dependence. The azimuthally varying data were then fitted with an equation relating phase velocity to its dependence on azimuth to azimuthal anisotropy, namely the strength of anisotropy and the fast direction. This procedure is repeated for all potential sub-arrays to obtain phase velocity measurements covering the contiguous U.S. We generate maps of the azimuthally averaged phase velocity for frequencies between 10-22 mHz for most of the contiguous U.S. that we compare to the statistically averaged phase velocity maps. Our resulting phase velocity maps are compared with previous studies to validate our methods.

Over the past few decades, online schools have witnessed exponential growth. Nevertheless, the body of research on their effectiveness and viability as educational options for students with disabilities, as well as their offerings for diverse learners, remains surprisingly sparse. This deficiency in evidence is noteworthy, especially as the demand for flexible education options continues to rise. In response to this gap, the present study conducts a literature review, which underscores the scarcity of empirical data detailing the specific attributes and benefits that online schools can provide. Despite the limited existing literature, our review identifies some distinctive factors, particularly related to the integration of the "5 Cs" learner control, flexible and rigorous curriculum, safe climate, caring community, and connection to students as individuals and their future goals. The focus of this study is to answer the following question: What are the unique experiences, perspectives, and challenges of SWD and their parents in an online charter school compared to their previous experiences in traditional brick-and-mortar public schools (TBMPS), encompassing aspects such as choice of online schooling, collaboration and communication, inclusivity, flexibility, parental involvement, and teacher support?This study is designed as a qualitative research endeavor, employing in-depth interviews with up to 25 families and their students with disabilities. The objective of this research is to gain a comprehensive understanding of the experiences of SWD within the online school environment. Furthermore, the study seeks to analyze these experiences through the framework of the "5C" design, shedding light on the key components of successful online education. By doing so, this research aims to contribute to a more nuanced understanding of the potential of online schooling, particularly for SWD, and provide actionable insights for educators and policymakers seeking to improve and adapt online educational programs to better serve diverse learners. Ultimately, this investigation addresses the pressing need to explore the untapped potential of online schools, with an emphasis on fostering an inclusive and effective learning environment

Marine ecosystems are facing various threats, from population declines to diseases that impact their overall health. My dissertation investigates the complex interplay between microbiomes, pathogens, and environmental conditions in three distinct marine organisms: Olympia oysters (Ostrea lurida) in the Puget Sound, Pacific oysters (Crassostrea gigas) in San Diego Bay exposed to OsHV-1 SDB µvar, and white abalone (Haliotis sorenseni) afflicted by Abalone Withering Syndrome. The goal of combining and comparing these systems is to elucidate the crucial role of microbiota in understanding ecosystem and host health, including microbes’ response to environmental variables and their interaction with pathogens. The first chapter focuses on the Olympia oyster, a native species in the Puget Sound that has experienced a substantial population crash. To assess the impact of eelgrass habitat and geographical location on oyster microbiomes, Olympia oysters from a single parental family were deployed at multiple sites, both within and outside eelgrass (Zostera marina) beds. Using 16S rRNA gene amplicon sequencing, I demonstrate that gut-associated bacteria differ significantly from the surrounding environment. Regional differences in gut microbiota are associated with the oyster survival rates at different sites after two months of field exposure. However, eelgrass habitat does not influence microbiome diversity significantly. This research highlights the importance of understanding the specific bacterial dynamics associated with oyster physiology and survival rates in the Puget Sound. In the second chapter, I explore the OsHV-1 SDB µvar, a virus threatening oyster aquaculture globally, with a focus on its microvariant in San Diego Bay. The study investigates the influence of temperature on OsHV-1 SDB µvar infectivity. All microvariants of this virus exhibit limited replication and are unable to induce oyster mortality at lower water temperatures. Through experimental infections of hatchery-raised oysters at temperatures ranging from 15 to 24°C, I found that no oysters died at 15°C but most exposed oysters died above 18°C. The infection took hold faster at 21 and 24°C compared to 18°C. As oysters are often immunocompromised by this viral infection, I also chose to focus on the potential contribution of secondary bacterial infections to the disease. The microbiome of healthy, sick and dead oysters was compared using 16S rRNA gene amplicon sequencing to determine how the microbiome is disrupted by infection and which bacteria may be responsible for further progression of the disease. There is a clear shift in microbiome composition and decreases in evenness following infection with OsHV-1 SDB µvar. The third chapter centers on Abalone Withering Syndrome, characterized by the intracellular parasite Candidatus Xenohaliotis californiensus (CaXc) which disrupts gut morphology leading to starvation and possible death. Investigating the microbiome in endangered white abalone exposed to CaXc over an 11-month period reveals dynamic variations in the fecal microbiome and its distinctiveness from the internal tissue microbiomes. CaXc exposure notably impacts the anterior region of the digestive tract more than the distal tissues and feces, sometimes representing up to 99% relative abundance in the post esophagus samples. This comprehensive analysis incorporates qPCR to quantify pathogen loads over time and feces and in internal tissues. The pathogen is detected after 5 months of exposure and is most abundant in the post-esophagus tissue. The samples with the highest relative abundance of the pathogen were also shotgun sequenced to generate whole genome assemblies of bacteria. This led to the novel assembly of a 90% complete genome for CaXc, which is deposited in a public database. To pair these data with a more holistic understanding of the impact of this pathogen, RNA sequencing data was analyzed for differential gene expression patterns between exposed and unexposed abalone. While functional annotation and prediction was poor on the de novo assembled transcriptome, clear differences exist in gene-level response to CaXc between post esophagus and digestive gland tissue.

RNA-binding proteins (RBPs) regulate post-transcriptional gene expression, influencing key cancer pathways. Understanding how RBPs control these processes can uncover new treatment strategies, which is crucial for cancers that lack effective targeted therapies such as triple negative breast cancer (TNBC). Here, we employ dual in vitro and in vivo CRISPR/Cas9 screens to investigate the role of RBPs in TNBC, revealing the poly(U)-binding splicing factor 60 (PUF60) as a key modulator of TNBC cell survival. Integrated eCLIP and RNA-sequencing identifies that PUF60 drives exon inclusion within proliferation-associated transcripts that, when mis-spliced, induce cell cycle arrest and DNA damage. Furthermore, disrupting PUF60 splice site activity via a substitution in its RNA-binding domain causes widespread exon skipping, leading to the downregulation of proliferation-associated mRNAs and inducing apoptosis in TNBC cells. We demonstrate that this RNA-binding disruption inhibits TNBC cell proliferation and shrinks tumor xenografts, highlighting the molecular mechanism through which PUF60 supports cancer progression. Our work demonstrates functional in vivo screening of RBPs as an effective strategy for identifying unexpected cancer regulators. Here, we reveal a crucial role for PUF60-mediated splicing activity in supporting oncogenic proliferation rates and highlight its potential as a therapeutic target in triple negative breast cancer.

Profiling binding repertoires of RNA binding proteins (RBPs) has uncovered how their intricate regulation shapes cellular function in health and disease. However many cell-type specific RBPs remain uncharacterized, especially within the CNS. We integrated our catalog of RBPs with cell-type annotations to identify microglial-specific RBPs and highlight ARID5A, an RBP previously implicated in peripheral immune cell RNA regulation, but not in microglia. We present integrated multi-omics analyses of ARID5A's RNA, DNA, and protein interactions, and leverage high-throughput sequencing and functional assays to uncover its RNA-mediated regulation of microglial functions. We show ARID5A binds RNA methylation sites, and regulates splicing and translation of transcripts integral to microglial functions. We found ARID5A modulates cytokine secretion, lysosome activity, and iron accumulation, as well as sensitivity to ferroptosis in microglia and co-cultured neurons. We illustrate the neuroprotective potential of ARID5A modulation, in which knockdown restored dysregulated functions in TREM2 mutant microglia. Our results not only emphasize the importance of profiling cell-type specific RBPs, but also highlight the potential of leveraging RBP-RNA interactions to restore dysregulated functions in neurodegeneration.

Deep learning (DL)’s dramatic rise in popularity across the domain sciences and industry has been accompanied by a correspondingly aggressive increase in the scale and computational complexity of DL workloads. In order to adopt state-of-the-art techniques, practitioners must wrestle with systems challenges of performance, cost, and scalability. In this dissertation, we identify the need for orchestration systems, which ease scaling burdens across the DL lifecycle through holistic, workload-aware optimizations. Drawing on both established techniques from data management research and new bespoke algorithms, we build practical orchestration engines to optimize three common DL workloads in the large-scale setting: model selection, data processing, and high-throughput serving. Our systems — which exploit workload- and context- specific opportunities — address a new layer of the large-scale DL optimization stack, more granular than current cluster managers and data systems, but still abstracted away low-level kernel & compiler optimizations. Empirical evaluations show that our orchestration techniques and systems can accelerate large-scale DL workloads by a large margin, even in complex, real-world settings. Our approach introduces a new technical lens, unifying systems, databases, and DL research, ultimately focused on democratizing and amplifying state-of-the-art DL innovations. Some of the systems proposed in this dissertation have already been adopted in production-scale industry pipelines, demonstrating the value of such orchestration optimizers for real-world DL.

The performance gains achieved by increasingly wide adoption of composite materials in the aerospace sector remain challenged by failure modes that are not always well understood. Implementing a structural health monitoring (SHM) system enables real-time monitoring of a structure, allowing for continuous assessment of its current condition.

This work proposes a comprehensive optimal fiber optic sensor placement framework for structural health monitoring applications. The framework is applied to an aircraft’s wing spar entirely made of composite materials. The damage of interest is debonding between laminates, which may cause local buckling that leads to reduced structural load-carrying capabilities. The development and validation of a high-fidelity finite element (FE) model that is used as a synthetic data generator are presented in detail. The inputs to the model are loads and debonding damage parameters (size and location), and the outputs are uniaxial strain measurements and buckling eigenvalues. Then, this work describes “run time” surrogate models created using different machine learning methods to overcome the high computational costs of each run of the physics-based model.

Bayesian inference is used to estimate the damage parameters given strain measured at candidate sensor locations. These estimations are used to assess damage criticality, which is linked to buckling eigenvalues, and transformed into decisions. Bayesian optimization is used to select the candidates that minimize a utility function that considers the costs associated with making a certain decision plus the costs of acquiring and installing the SHM hardware (sensors, data acquisition system, etc.). The goal of the optimization is to find the sensor set that provides the lowest cost. The resulting optimal sensor configuration is presented, consisting of the number of sensors to be deployed and their respective locations. Finally, this work analyzes the performance achieved by several different cost functions, highlighting the importance of defining an objective function that reflects the goal of the SHM system, demonstrated to an aerospace application.

Metabolic Syndrome (MetS) affects one in three adults in the US and is characterized by abdominal obesity, high blood pressure, high blood sugar, high triglycerides, and low HDL cholesterol, increasing the risk of cardiovascular diseases and diabetes. We explored differences in neighborhood conditions and psychosocial factors between men and women across low and high socioeconomic status (SES) groups, and how these differences contribute to the higher risk of MetS in low SES individuals. Our sample included 4,191 individuals aged 45-84 from diverse racial and ethnic backgrounds; 3,005 of them were classified as low or high SES for further analyses. We examined the association of MetS with SES for women and men separately using separate Cox proportional hazards models for each group. The mean age was 61 years, 49.2% were women, 43% were White, 9.4% had no health insurance, and 15.2% had less than a high school education. The overall incidence of MetS was 33% for both low and high SES individuals. After adjusting for confounders, low SES women had a 66% higher risk of developing MetS compared to high SES women (HR 1.66, 95% CI 1.38 to 2.00). Similarly, low SES men had a 66% higher risk of developing MetS compared to high SES men (HR 1.66, 95% CI 1.36 to 2.02). Contrary to expectations, adjusting for these variables did not attenuate the risk, and significant disparities were observed for both men and women across SES groups. More research is needed to understand the mechanisms linking low SES with MetS incidence to address these SES disparities.