UC Riverside

Cancer is the second leading cause of global mortality. Accumulating evidence strongly suggests that aberrant alterations in signaling cascades can lead to uncontrolled cell growth, driving the progression of this disease. While the promotion of cancer cell growth by hyperglycemia has been extensively studied, the comprehensive mechanism underlying this relationship remains incompletely elucidated. In the current study, we present findings that high glucose levels (HG) indeed bolster DNA synthesis and cell cycle progression, fostering tumor cell growth, as revealed by genome-wide analyses. Moreover, our investigation identified E2F family as the central transcription factor responsible for the adaptive response induced by HG. The use of the pan-E2F inhibitor HLM006474 and shRNA-mediated E2F1 depletion, effectively counteracted HG-triggered DNA synthesis and cell growth. Furthermore, we establish that HG amplifies Rb1 phosphorylation, directly contributing to the activation of E2F1. Inhibition of Rb1 phosphorylation using the CDK2/4/6 inhibitor compound PF-3600 substantially mitigates the E2F1-mediated transcriptional activation of downstream DNA replication genes. Notably, among the E2F1 target genes induced by HG, RRM2 plays an essential role in the nucleotide synthesis by generating essential dNTPs required for DNA replication. Our investigation reveals that HG promotes intracellular dNTP levels in an E2F1-RRM2-dependent manner, closely associated with increased DNA synthesis and subsequent cancer cell growth. The RNR inhibitor Triapine effectively impedes RRM2-mediated elevation of intracellular dATP and dGTP, and DNA synthesis. To gain more insights into the metabolic alterations occurring in HG-treated cancer cells, we carried out the untargeted metabolomic profiling. This analysis not only reaffirms the activation of glycolysis in HG-treated cells, as evidenced by the elevated levels of lactic acid, but also illustrates noteworthy enhancements in the levels of sucrose and fructose in the polyol pathway. These findings provide another facet of how cancer cells utilize excess glucose to fuel their proliferation. Collectively, our findings decipher the oncogenic signaling and metabolic connections between hyperglycemia and cancer cell proliferation, mediated through the Rb1/E2F and polyol pathways. This study provides a detailed molecular mechanism that sheds light on how hyperglycemia directs tumor cells to enhanced cell cycle progression and cell proliferation.

Mitochondria are important subcellular compartments, crucial for energy production, metabolism, and cell signaling. Mitochondrial dysfunction is known to cause nearly 200 mitochondrial disorders and has been associated with aging and a variety of human diseases. Mitochondrial DNA (mtDNA) encodes 37 genes, including 13 proteins and a set of tRNA and rRNA. mtDNA is constantly threatened by chemical and physical assaults. Because mitochondria have limited DNA repair pathways, mtDNA damage accumulates and occurs at a higher level compared to nuclear DNA. Abasic (AP) sites are abundant DNA lesions that can be generated from various pathways, including base excision repair (BER). AP sites are highly reactive and can form secondary DNA adducts, DNA-interstrand cross-links (ICLs), and DNA-protein cross-links (DPCs). My dissertation project exploits the chemistry of AP sites and develops methods to explore biological processes pertinent to AP sites. First, I developed a mass spectrometry-based method to identify the cross-linked amino acid residues in DNA-protein cross-links. I designed DNA substrates with ribonucleotides (rNMPs), which provide chemical-labile sites for DNA strand cleavage reactions and produce structurally defined DPCs. Also, I developed a program (AP_CrosslinkFinder) to accelerate data analysis. The method was applied to identify the cross-linking amino acid residues in DPCs derived from mitochondrial transcription factor A (TFAM). Second, I developed a method to prepare model ICLs using rNMP-containing DNA with a nucleotide analog 2-aminopurine. The alkaline lability of rNMP enables the generation of strand breaks at specific sites. AP sites react with 2-aminopurine with high yield and high rate. This method provides a simple and straightforward tool for investigating the impact of ICLs during the repair process. Third, I investigated the DNA terminal structures generated in TFAM-catalyzed AP-DNA strand cleavage. Quantification of reaction rates in the presence of biological amines and thiols demonstrates that GSH competes with TFAM for AP site strand breaks, suggesting a possible strategy to limit the formation of DPC and control the strand break terminus in cells. Removal of DNA terminal modifications by relevant DNA repair enzymes was also evaluated. Together, results from my dissertation provide insights into the complexity of AP site chemistry with important biological implications.

The reliability of Very Large Scale Integration (VLSI) circuits is crucial in modern electronic devices. VLSI circuits, which contain millions of transistors, are vulnerable to a variety of reliability issues such as electromigration (EM), time-dependent dielectric breakdown (TDDB), and temperature variation. These issues can lead to circuit failure and reduce the lifetime of electronic devices. Traditionally, VLSI reliability analysis and prediction have been performed using physics-based models and simulators. These models, however, are computationally intensive and can be time-consuming to run. In recent years, machine learning (ML) techniques have been used to predict and diagnose reliability issues in VLSI circuits. This thesis presents an in-depth study of machine learning techniques applied to EM analysis, post-silicon thermal map estimation, and electrostatics analysis. Specifically, the first segment proposes two data-driven ML methods for the fast prediction of transient EM stress of general interconnects in VLSI circuits. The traditional numerical partial differential equation (PDE) problem is treated as an image processing or graph aggregation problem which yields considerable speedup with acceptable accuracy. However, these methods are still supervised learning approaches, requiring extensive training data generated from numerical solvers. Therefore, the second segment proposes a hierarchical physics-informed neural networks (PINN) based method for EM analysis. This approach leverages PINN, which is trained mainly by physics laws with minimal labeled data requirements. The hierarchical nature of interconnects is leveraged, and the entire interconnect tree is solved step by step. Temperature variation has always been problematic in VLSI circuits, as reliability degrades drastically as temperature varies. The third segment presents a real-time thermal map estimation method for commercial VLSI circuits. This approach treats thermal modeling as an image-generation task using generative neural networks (GANs), producing tool-accurate thermal map estimations. Electrostatics analysis is an essential step for analyzing TDDB, an important failure mechanism for interconnects. Lastly, the fourth segment presents a PINN-based 2D electric field analysis method. This approach eliminates the heavy dependence of finite element methods (FEM) used in traditional TDDB analysis and leads to orders of magnitude of speedup.

Reputation consists of what is said or believed about a person by those who know the person. Our reputation may affect the opportunities available to us and the way others treat us. The current research reports a series of six studies (total N = 2116) used to develop and validate a self-report questionnaire assessing people’s interest in their reputations and their efforts to manage their reputations. The first step in item development relied on open-ended responses to questions relating to reputation interest and management. Using a rational/factor-analytic approach and multiple samples, a two-scale (Concern, Knowledge) Reputation Interest measure and a four-scale (Discretion, Objecting, Masquerading, Support Seeking) Reputation Management measure were produced. Theory-relevant relationships between these multidimensional reputation scales and Big Five personality traits, well-being, and various interpersonal personality characteristics demonstrated validity of the reputation scales. For example, scores on the Masquerading scale (measuring a tendency to present a false appearance of oneself) were positively associated with measures of neuroticism, manipulativeness, approval motivation, and dependency, and negatively associated with a measure of conscientiousness. In the final study, agreement of informant reports and target reports on respective dimensions of reputation interest and management contributed to demonstrating validity of the multidimensional reputation scales. These informant reports also provided information regarding the reputation of target participants. For example, targets who score high on the Knowledge scale (measuring the degree to which one believes to have an awareness of one’s reputation) are reputed by informants to have a higher level of extraversion, agreeableness, openness, and peer acceptance. The value of the reputation scales in providing a broader personological picture is discussed as are other potential benefits of the scales.

The Mandelbrot set, a fractal well known in the world of mathematics, exhibits many interesting properties for music composition. In fact, when read as a visual score from top-left to bottom-right, this fractal not only creates form through its horizontal symmetry, but has clear areas of tension and release ideal for creating exciting music. As the Mandelbrot set is visually beautiful, this project focuses on creating audiovisual works of art using data sets from the main set, zooms of the set, and both mild and extreme functional alterations. While basic sonifications of the set have been explored by other artists before, this project seeks to create a multi-movement work where each sonic and visual element is controlled by or directly relates to the Mandelbrot set. This raises a number of interesting questions, most importantly: How can one translate math into art in a way that there is a clear, meaningful connection between the art and the source? While answers may vary, the most effective answer for this project was to focus on recreating my personal musical language in the form of a mathematical formula, through which data points from the Mandelbrot set are processed.

Urbanization is rapidly changing the structure and function of freshwater ecosystems across the planet. Southern California is currently experiencing an advanced urban stream syndrome regime due to the dense human population. Urban alterations have resulted in changes to rivers and streams that include physical alterations, effluent discharge that changes hydrology and water chemistry, the introduction of non-native and invasive species that alter biotic filters, and management of these systems to preserve threatened and endangered species. This suite of changes results in a patchwork landscape for species within a city. I examine the heterogeneity that exists within a highly urbanized river to better understand urban heterogeneity, its impacts on freshwater trophic structure, and species foraging preference. In chapter one I present the results of monthly habitat and benthic community surveys across an urban gradient containing three wastewater treatment plants. I found that these impacts did not have consistent impacts on habitat or freshwater benthic communities across the studied gradient and that certain habitat variables had strong impacts on diatom and macroinvertebrate species richness and density. My second chapter investigates the role of wastewater treatment plants on community trophic structure and invasive species diets across three wastewater discharge channels and the main stem of an urban river. I found that wastewater facilities had different impacts on nutrient enrichment, community trophic structure, and invasive species diets. These changes were not consistent between wastewater and main stem channels with trophic compression occurring in each. Chapter three examines how wastewater facilities impact a federally threatened species foraging preference within an urban river. We found that this species had a clear preference for the forage below one of the three wastewater discharge points and for forage in the main stem of the river away from wastewater inputs. This preference did not overlap with the species current distribution in the river. Throughout this dissertation, I demonstrate the level of urban heterogeneity that exists within this system, how the impacts of alterations had different impacts, and present opportunities to improve management of urban freshwater ecosystems.

Although sustainable thermal energy storage and conversion approaches are becoming increasingly popular, high-power requirements (~107 kWh) in certain energy applications such as spacecraft and jet propulsion, can only be accomplished through direct conversion of chemical to thermal energy through combustion. The current research objective in this field is to alleviate the kinetic and mass transfer limitations impeding rapid energy extraction from solid state fuels possessing high energy density. Part of the dissertation focuses on solving such challenges for boron particles and boron containing complexes, as elemental boron has the highest energy density. This dissertation provides mechanistic insights on selectively altering the thermochemical decomposition pathways of boron containing solid-state borane-ammonia complex with the help of a) ammonium-ion based oxidizing salts and b) polymeric carbonyl groups, to facilitate rapid and complete oxidation of boron in the borane-ammonia complex, which otherwise gets kinetically trapped into boron-nitrogen-hydrogen clusters. The dissertation also provides atomic scale understanding of the manipulation of boron oxide shell by implanting magnesium atoms, as a potential strategy to enhance reactivity of boron particles. Magnesium simultaneously reduces boron oxide creating dangling bonds on boron and induces a net tensile strain on the boron oxide surface, which respectively enhances the adsorption and diffusion rate of oxygen through the boron oxide shell, collectively enhancing the oxidation rate of boron. Significant part of the dissertation also focuses on tuning the reactivity of composites containing aluminum, another high energy density and widely explored solid fuel, by alteration of micro- and meso-structural features, as well as the incorporation of microwave absorbing sensitizers in aluminum-based composites for spatial confinement of the ignition zone when activated using microwave radiation. As a major breakthrough, this dissertation also demonstrates that certain non-flammable condensed phase materials with high energy density such as imidazolinium-ionic liquids can be electrochemically made inflammable and their flame can be dynamically extinguished simply by removing the voltage bias driving the electrochemical reactions. Since most energy dense fuels pose the danger of unintended fire and explosion, this concept paves the path for the development of potentially ‘safe’ fuels, thereby opening multiple avenues for future research.

Anthropogenic climate change represents one of the most serious threats to ecosystems in the 21st century. As temperatures increase, and precipitation patterns are altered, species need to respond to living in increasingly arid environments. The most noticeable responses to changing climate is for populations to shift spatially, typically upward in elevation and latitude, and phenologically, typically by becoming phenologically active earlier in the year. Variation in how individual organisms or populations respond to climate change can alter their ecological interactions. The timing of flowering is species specific, and when and with whom a plant flowers adjacent to can impact their reproductive success. Between trophic levels, the synchronous phenology of flowering plants and pollinators is critical for both plant and pollinator reproductive success. Plant-plant and plant-pollinator phenological synchrony is at risk of deterioration due to aridification, potentially decreasing ecosystem functioning across the globe. While the bulk of previous research on this issue has been conducted in humid systems, plant- pollinator phenological synchrony has been in understudied in dryland ecosystems, which encompass over 40% of land globally. In the following chapters, I leverage natural history data along spatial and temporal gradients to determine the impacts of climatic variation on plant-plant and plant-pollinator phenological synchrony. I find evidence that plant-plant phenological synchrony is sensitive to changes in community composition. Plant-pollinator phenological synchrony decreases with increasing aridity at the community level, but some species are better suited to future aridification than others. My dissertation highlights the importance of understanding phenological synchrony in dryland ecosystems using analytical techniques specifically suited to the stochastic nature of climate change in these systems.

We have developed a post-processing tagging technique to model globular clusters (GCs) in cosmological hydrodynamical simulations. We have applied our method to the Illustris and TNG50 simulations to study several aspects of GCs in galaxy groups and clusters, regimes where no other theoretical technique is available to link galaxies, halos and GCs. We find that GCs are good tracers of dark matter–both in terms of their radial distribution and shape to trace the host halos in groups and clusters, but also through their kinematics to constrain the dynamical mass of dwarf galaxies. We have used our catalogs to establish one-to-one comparison to observational determinations of dark matter mass in dwarfs, finding that while systems with more than 10 GC may recover the right dark matter mass content via Jeans modeling and other derived mass estimators, for dwarfs with less than 10 GCs, assumptions in the prior and different methods to estimate the velocity dispersion may heavily bias the ability to infer dark matter mass from kinematics. We find that dwarf galaxies are consistent with populating an extrapolation of a single power-law relation between GC mass and halo mass observed for more massive systems, at least all the way down to dwarfs comparable to dSphs in the Local Group (M∗ = 5×106M⊙). Lastly, we explore the GC systems of the set of morphologically defined ultra-diffuse galaxies (UDGs) within the TNG50 simulation. Observationally, the kinematics of the GC systems of UDGs show a large diversity, with systems ranging from apparent “failed galaxies” living in overly- massive dark matter halos to the opposite extreme, where UDGs are seemingly dark-matter free. We use our GC catalog to demonstrate that much of this diversity, in particular towards low GC numbers, might arise as a combination of a low number of dynamical tracers coupled to ongoing tidal disruption—in agreement with evidence of stellar streams in some of the UDGs with low velocity dispersion—as well as contamination from GCs in the intracluster medium.

Deep neural networks have demonstrated remarkable efficacy across a wide range of tasks, yet they face a significant limitation in their ability to adapt to distributional shifts. In contrast, humans possess an inherent adaptability, effortlessly adjusting to changes in data distributions and modifying task strategies to accommodate environmental variations. To fully harness the potential of deep learning models and enhance their practical applicability, it is crucial to impart robustness to distributional shifts. This dissertation addresses this need by presenting algorithms to empower deep learning models with the capacity to seamlessly navigate diverse forms of distributional shifts.

The dissertation encompasses four significant contributions. First, we explore the adaptation of a person re-identification model trained on labeled data from a single camera to other cameras in the network using only unlabeled data. By optimizing temporal consistency across frames in unlabeled videos, the model acquires generalizable representations. Second, we address the adaptation of 2D human pose estimation models to different imaging conditions, achieving adaptation through pre-trained models and unlabeled data from the target domain. Leveraging a pre-built human pose prior that captures plausible human poses, labeled data becomes unnecessary for the adaptation process.

Expanding the concept of adaptation beyond static tasks, we proceed to tackle sequential decision-making problems. It demonstrates how imitation learning can be executed when expert demonstrations originate from domains with distinct morphologies compared to the learning agent. By utilizing cyclic state transformation consistency and value function consistency, a transformation function is learned to render demonstrations comprehensible to the agent.

Finally, we shift focus towards adapting to user constraints, a critical aspect of deep learning model adaptability. It addresses the challenge of adapting multi-task models to changing user preferences by introducing a hypernetwork controller capable of dynamically modifying model architecture and weights without necessitating re-training.

By bridging the gap between human adaptability and the limitations of current models, this dissertation paves the way for deep learning to become more versatile and applicable in real-world scenarios, unlocking its full potential across various domains.