Open scholarship at Cornell
eCommons is a service of Cornell University Library that provides long-term access to a broad range of Cornell-related digital content of enduring value. Learn more about eCommons.
Make a deposit
Submit your document, research paper, image, data, code, and more into Cornell's digital repository. Learn how to submit.
Recent Submissions
this is the rug i wanted
johns, erica (CUL, 2024-04-26)
what is the difference between an abstract and a description?
DISSECTING HETEROGENEITY OF IMMUNE CELLS: FROM DIVERGENT PROPERTIES OF NAÏVE CD8+ T CELL SUBSETS TO MONOCYTE DYSREGULATION IN ME/CFS
Zhu, Hongya (2023-12)
Biological systems are inherently heterogeneous. In vertebrates, the immune system protects the body with diverse cells of specialized functions. CD8+ T cells are a major component of adaptive immunity, contributing to pathogen clearance and tumor control. Naïve CD8+ T cells are a heterogeneous population with subsets of distinct kinetics and functions upon activation. It is increasingly clear that their divergent post-stimulation fates can be established at naïve stage, but gene expression programs differentiating the naïve subsets remain to be elucidated. Here, I analyzed a diverse set of RNA-seq and ATAC-seq profiles in naïve CD8+ T cell subsets, where extensive differences in gene expression and chromatin accessibility were observed. Leveraging those profiles, I used a powerful network inference algorithm, Inferelator, to construct a transcriptional regulatory network in naïve CD8+ T cells. Key TFs promoting each naïve subset were identified, which not only included known TFs, but also implicated novel roles for additional factors. Taken together, this work investigated transcriptional control on naïve CD8+ T cell subsets, revealing regulatory circuits that preprogram their distinct fates. In immune-related diseases, dysregulation of immune cells may display to different degrees across patient cohorts, introducing additional layers of heterogeneity that are important to decipher for elucidating disease mechanisms. Myalgic encephalomyelitis/chronic fatigue syndrome (ME/CFS) is a serious and poorly understood disease, characterized by symptom exacerbation following exertion. To understand immune dysregulation in ME/CFS, we used single-cell RNA-seq to examine immune cells in patients and controls at baseline and after symptom provocation. Monocyte dysregulation was identified as a prominent feature of immune dysregulation in ME/CFS, with patterns suggestive of inappropriate differentiation and migration to tissue, both at baseline and after symptom provocation. Importantly, mixed monocyte populations, including diseased and more normal cells, were identified in patients, and the fraction of diseased cells correlated with metrics of disease severity. Taken together, this work investigated immune dysregulation in ME/CFS at single cell resolution, and identified heterogeneous monocyte dysregulation in ME/CFS patients. Overall, this dissertation presents two projects that examine heterogeneity of immune cells: identifying regulatory circuits differentiating naïve CD8+ T cell subsets, and unraveling immune dysregulation in ME/CFS.
ANTI-CANCER ACTIVITIES OF OLEANOLIC ACID AGAINST TRIPLE-NEGATIVE BREAST CANCER CELLS: MECHANISMS OF ACTION
Zhao, WeiYang (2023-12)
Breast cancer, notably Triple-negative breast cancer (TNBC), a subtype lacking expression of estrogen receptor (ER), progesterone receptor (PR), and human epidermal growth factor receptor II (HER2), remains a significant global health issue due to its poor prognosis and high invasiveness. This study investigates oleanolic acid (OA), a compound that exhibits a broad spectrum of biological activities, such as anti-inflammatory, hepatoprotective, analgesic, cardiotonic, sedative, and anticancer properties. Numerous investigations have explored the potential health advantages of OA derived from natural sources, identifying it as a promising candidate for the development of novel preventative and therapeutic agents for various ailments. Additionally, the abundance of OA in plant-based foods renders it an appealing target for dietary interventions and the development of functional foods. However, the anti-cancer profile of OA against TNBC cells has not been fully elucidated. We discovered that OA could inhibit TNBC MDA-MB-231 cell proliferation and induce cell cycle arrest. OA's effect is realized through the downregulation of cyclin D1, CDK4, and p-cdc25c activities, analyzed via methylene blue assay, flow cytometry, and Western Blot. Moreover, OA was found to disrupt EGFR/Ras/ PI3K/ AKT/mTOR and EGFR/ERK1/2/GSK-3β pathways, thereby modulating cell cycle checkpoint proteins p53 and p21Cip1/Waf1. We also noted OA's ability to counteract EGF-induced protein upregulation and to promote cell cycle checkpoint proteins. Our study further elucidates the pro-apoptotic role of OA. The compound's apoptotic influence, marked by increased BAD and cytochrome c expression, was observed under various dosages. OA also induced a decrease in ATP production, signifying a metabolic shift, and downregulated mitochondrial-associated BCL-2 family proteins, potentially affecting apoptosis initiation markers. OA additionally modulated key apoptotic and cell cycle regulators in an EGF-stimulated cell model, leading to upregulated expressions of p21Cip1/Waf1, p53, caspase-9 and caspase-3, and a decrease in p-JNK expression. Lastly, our investigation into OA's anti-metastatic properties reveals its ability to downregulate matrix metalloproteinases MMP-2 and MMP-9, decrease the activity of released MMP-2 by modified zymography, and the reduced expression of migration marker VEGF, regulated by the c-Jun transcription factor. These findings suggest that OA's inhibitory effect on metastasis occurs via EGFR/p38 MAPK/JAK3/STAT-3, EGFR/ERK/STAT-3, and EGFR/JNK/c-Jun signaling pathways. Results of this study illuminate OA's potential therapeutic benefits in managing TNBC proliferation, apoptosis, and metastasis, thereby offering a promising avenue for the development of more effective treatment strategies from the perspective of diet and cancer.
RENEWABLE TRANSITION PLANNING AND OPERATIONS SCHEDULING FOR SMART ENERGY SYSTEMS USING OPTIMIZATION AND MACHINE LEARNING
Zhao, Ning (2023-12)
This dissertation deals with the transition planning and operations scheduling of smart energy systems using optimization and machine learning, aiming to address challenges in accommodating renewable energy penetration. This dissertation covers four aspects, including the development of decarbonization pathways for regional energy transition, incentive policy design for the adoption of renewable energy, waste-to-energy supply chain optimization, and power systems operations with intermittent renewable energy. The first research aspect includes three distinct research projects. In the first related project, we propose a bottom-up optimization modeling framework for energy transitions, which bridges the power sector and other energy sectors represented by space heating. In the second related project, we develop a bottom-up data-driven multistage adaptive robust optimization framework to investigate energy transition pathways under uncertainty. We apply affine decision rules to overcome the computational intractability. Machine learning techniques are applied for constructing data-driven uncertainty sets to avoid over-conservation. In the third related project, we propose a multi-scale bottom-up renewable electricity transition optimization framework to simultaneously address yearly systems design decisions and hourly operations decisions. A novel machine learning-based approach is used to reduce computational demand. For the second research aim, we propose an optimization framework based on single-leader-multiple-follower Stackelberg game to promote bioenergy generation. We formulate a bi-criterion mixed-integer bilevel fractional programming problem, and a tailored global optimization algorithm integrating a parametric algorithm and a projection-based reformulation and decomposition algorithm is developed for solving. For the third research aim, we develop a life-cycle optimization framework for poultry waste supply chain considering fast and slow pyrolysis technologies, which is formulated into multiobjective mixed-integer fractional linear programming. The fourth research aspect includes two distinct projects. In the first one, we propose a machine learning-based two-stage adaptive robust optimization framework with data-driven disjunctive uncertainty sets for volatile renewable generation to address renewable energy-induced disjunctive uncertainties in power systems operations. To facilitate the solution process, a tailored decomposition-based optimization algorithm is developed. In the second project, we develop an optimal power flow framework to address the operations of the future zero-carbon grid, aiming to involve hydrogen-based long-term energy storage to tackle the seasonal load shedding and transmission line congestion issues of deeply decarbonized power systems while capturing the variety of future climate scenarios, the topology, and operational requirements of the power systems.
Co-Design of Binarized Deep Learning
Zhang, Yichi (2023-12)
The 2010s oversees the explosive growth of deep learning and its profound impact beyond. This technology has fundamentally revolutionized the paradigm of pattern recognition and decision making in a wide range of applications and is believed to be one step further towards machine intelligence. One common trend observed by researchers over the years is the persistent scaling of model, data, and hardware. In fact, today's large-scale deep learning models have hundreds of billions of parameters and septillions of floating-point operations, trained on thousands of chips using trillions of tokens. Consequently, deploying such models at scale keeps stretching the limit of today's supercomputers. The core problem studied in this dissertation is quantizing deep neural networks to one bit. Quantization reduces the number of bits in numerical representations. Quantizing matrix multiplications, the most compute-intensive part of deep neural networks, can substantially reduce their energy footprint and improve their performance. One-bit quantization, referred to as binarization, embraces its extreme benefit. Although there is a lack of theoretical evidence, binarization empirically suffers from degraded model qualities. This dissertation explores the possibility of co-designing both high-quality and high-performance binarized neural networks (BNNs). In part one, it introduces the precision gating (PG) technique that leverages additional sparse binary matrix multiplications to improve the accuracy of existing BNN architectures. The result BNN with PG, named as FracBNN, for the first time achieves the same accuracy as a well-known compact floating-point network MobileNetV2 and is capable of performing real-time inference on an embedded FPGA. In part two, it introduces the design methodology of PokeBNN, a novel vision BNN architecture that establishes a new Pareto-state-of-the-art in terms of the accuracy-efficiency trade-off. In part three, it further explores binarized Transformer-based language models. The one-bit weight-binarized Transformer demonstrates no quality loss on WMT De-En translation dataset and a similar scaling trend as its floating-point counterpart when evaluated on Google Translate's production-scale dataset. Finally, the dissertation presents a comprehensive and automated model deployment framework for quantized deep neural networks. With one line of code addition, the framework can deploy mixed-precision models on CPUs and demonstrate a practical acceleration.