80264 research outputs found
Sort by
Using Deep Learning to Extract Multicellular Aggregation Features of Myxococcus xanthus
No full textSelf-organization is a fundamental biological process whereby local interactions among individual elements give rise to emergent global behaviors and complex patterns. While widely observed, the genetic determinants and biophysical mechanisms underlying self-organization—especially in prokaryotic multicellularity—remain incompletely understood. The bacterium Myxococcus xanthus, known for its intricate aggregation and fruiting body formation under nutrient deprivation, offers a powerful model for studying these dynamics.
This thesis addresses both methodological and analytical challenges in studying M. xanthus development. To overcome the limitations of traditional imaging (e.g., difficulty in segmenting dense aggregates from phase-contrast alone), we developed a generative adversarial network (GAN) that synthesizes fluorescence microscopy images from phase-contrast images. This approach combines the ease of phase-contrast acquisition with the quantitative advantages of fluorescence imaging, improving aggregate segmentation and enabling accurate analysis of multicellular patterns such as rippling waves.
We also introduce a deep learning framework for quantitative phenotypic analysis, integrating ResNet and StyleGAN2 into a Variational AutoEncoders (VAEs), and using Siamese architectures as the similarity metric. This pipeline transforms high-resolution microscopy data into low-dimensional phenotypic feature vectors. Human evaluations confirmed the model captures phenotypic diversity, enabling precise strain-level comparison of developmental dynamics.
Altogether, this work demonstrates how deep learning can uncover genotype–phenotype relationships in bacterial self-organization and offers broadly applicable tools for the study of emergent behavior in biological systems
Analog Front End for Miniaturized Neural Implants
No full textThis thesis presents a miniaturized neural implant with magnetoelectric power and wireless communication. The analog front end (AFE) design consists of a low-noise amplifier (LNA), a noise-shaping SAR ADC, and a digital post-processing block. Conventionally, the first stage is a high-gain LNA to maximize energy efficiency. However, the motion/stimulation artifacts cause the LNA saturation issue. Recently, more papers have presented ADC-only structures to address this issue. Although ADC-only structures can achieve high energy efficiency, the energy and area efficiency of the system are limited due to the high oversampling ratio (OSR). The proposed design achieves the lowest OSR with comparable noise performance and input range among recent works. For the implant’s wireless uplink communication, we propose adopting energy extraction techniques to enhance the SNR and increase the data rate. The implant SoC is fabricated using standard TSMC 180-nm CMOS technology. The SoC chip has an active area of 2.6 mm2, and the AFE has an active area of 0.372 mm2 with 6.16 µW power consumption. The AFE achieves 8µVrms noise at the AP band and 11.35 µVrms noise at the LFP band. The wireless uplink communication achieves a 17.73 kbps data rate with 0.9pJ/bit energy efficiency. The chip is validated through an in-vitro test demonstrating the LFP recording, wireless power transfer, and communication
Fort Bend County Community Health Assessment: Insights from the Greater Houston Community Panel
No full textGiven the relationship between financial resources and health, Fort Bend Health and Human Services enlisted UTHealth Houston School of Public Health to conduct a community health assessment of county residents. In the fall of 2024, the latter collaborated with the Kinder Institute for Urban Research on the assessment. This report provides a brief overview of the results, describing the current state of health and other community needs in the county
Electoral Incentives and Political Support for Anticorruption Reform: Evidence from Latin American Legislatures
No full textThis dissertation examines the conditions under which politicians advance anticorruption policies. Anticorruption policies rarely receive widespread political support because they can be costly and risky for politicians. Yet, politicians still propose and advocate for these reforms. Why? In my three-paper dissertation, I argue (and show) that legislators are strategic when it comes to anticorruption, seeking to sponsor policies that will get them votes without jeopardizing their careers and rents. I leverage original data on anticorruption bills introduced to Latin American legislatures, natural experiments, survey experiments, and interviews with politicians and activists to document the role of electoral incentives. Overall, results show that 1) anticorruption reform is possible under the right conditions, 2) legislators are responsive to electoral incentives but will seek to minimize the potential consequences of anticorruption policies, and 3) voters evaluate anticorruption efforts favorably, but certain contextual features make some appeals more credible. Legislators are more likely to sponsor anticorruption initiatives after high-profile corruption scandals, when they are members of the opposition and when they are up for reelection. Furthermore, legislators are more likely to sponsor punitive policies, which they believe are more popular and less likely to become law than non-punitive policies. Finally, evidence from a survey experiment in Mexico suggests that there is a mismatch between what politicians believe will get them votes and what voters prefer. While voters are more likely to view opposition party legislators sponsoring anticorruption bills more favorably than incumbents, they are not more likely to prefer punitive over non-punitive action, and they evaluate anticorruption policies more highly if they are sponsored in the absence of a scandal. Together, these papers contribute to our understanding of anticorruption policies, politicians' incentives to advance them, and voters' evaluations of these efforts
Efficient, Agile, and Flexible Network Resource Sharing With INT-enabled Decentralized Control
Get PDFWith such large-scale networks and massive endpoints, decentralized control is prominent in providing lightweight, scalable network resource allocation. Previous solutions achieve such decentralized control with congestion control systems in the network, but improving efficiency, agility, and flexibility for such control remains a significant question since the network resources are allocated as a black box with little visibility. However, for decades, congestion control (CC) protocols have struggled to gain visibility into networks' fine-scale hop-level congestion state. The key knob in this thesis is the in-band network telemetry (INT) signal, which is a recent capability of commodity switches where the traffic and endpoints can obtain certain telemetry information from the routers. With such fine-grained quantitative information provided by routers agilely, the potential of CC protocols has been largely expanded. Along with all the above merits of the INT signal, this thesis determines the best INT signals for CC and develops a novel decentralized protocol on endpoints to converge the whole network allocation towards a global goal.
Poseidon obtains the per-hop queueing delay signal from the INT and uses this signal to guide the rate update of each endpoint. designed a decentralized protocol around this quantitative signal, achieving agile bandwidth resource sharing with fast convergence towards the final allocation. For practical deployment, Poseidon also determines the INT format and the collecting mechanism for the delay signal with multiple vendors' standards. Then we realize that even with a small amount of queueing, the extra queueing delay can be a problem for latency-sensitive applications. With this observation, we design a novel INT signal --- arrival rate utilization --- to capture the network status accurately without any queueing. In C2L2, we propose a CC algorithm with much better efficiency, which achieves zero queueing while keeping all other good properties, such as high utilization, fast convergence, and fair allocation. Finally, a more flexible resource allocation policy, such as differentiated services, is required to support various cloud applications' needs. Thus, in Soze, we design a flexible INT-based resource allocation system that allows a custom policy on each individual endpoint, achieving agile and accurate weighted resource sharing
Novel Random Algorithms for Solving Nonograms Efficiently
Get PDFNonograms are a fun and intriguing puzzle game. Most human-made nonograms are relaxing and solvable relatively quickly using logical steps; however, solving the general case of nonograms is hard. General nonogram solving is an NP-complete problem, meaning it is at least as hard as the hardest NP problem, and that there does not currently exist a fast algorithm for finding a solution. This project creates randomized algorithms for solving nonograms that are expected time better than non-randomized algorithms
Investigation and Development of Methods to Produce Budding Yeast Transcriptomic Aging Data
Get PDFGiven the resources and proper environment, an isogenic S. cerevisiae population has the potential to propagate indefinitely. However, the individual yeast cells within a colony are not immortal and rejuvenation through budding is required for the continual production of youthful progeny. The rejuvenation process in yeast budding has been shown to be a victim of aging too, resulting in several known disparities between early and late daughter cells from the same mother cell. The transcriptomic signatures of sets of individual yeast daughter, or progeny, cells from throughout the entire lifespans of yeast mother, or progenitor, cells would reveal the changes that result from the age-induced breakdown of yeast budding rejuvenation and provide a meaningful measurement of biological aging. Using such data, it would be possible to construct a yeast single-cell transcriptomic aging landscape defining cellular biological age in an unbiased manner for yeast daughter cells. Further coupling the daughter series transcriptomes with those of yeast mother-daughter pairs would allow for the extension of the daughter cell aging landscape to a general yeast aging landscape which would detail the biological age of a yeast cell given its transcriptome. Currently, methods to capture single-cell yeast lineage data are lacking and need to be developed. As my thesis project, I developed and analyzed methods for plate-based single-cell and a bulk single-cell sequencing methods for budding yeast to be used for the establishment of a transcriptomic definition of cellular biological age at single-cell resolution which will further improve the utility of yeast as a eukaryotic cellular aging model organism
How Strategy, Situation, and Person Interact to Predict Adaptive Emotion Regulation: A Personalized Approach to the Science of Emotion Regulation
No full textEmotion regulation is pervasive in daily life and critical for well-being. There are calls to investigate emotion regulation adaptivity through the interaction of person, situation, and strategy. Recent work has begun investigating adaptive emotion regulation as a function of situation and strategy factors, where individuals vary in emotion regulation strategy use and efficacy. Additionally, evidence suggests that cultural values are an important individual characteristic that determines when and how emotion regulation strategies are used. However, there is a need to expand this work in order to incorporate more ecologically valid methodologies, assess a broader range of emotion regulation strategies,
apply this knowledge in emotion regulation training, determine longitudinal impacts of such training, and combine all three factors in one model to gain insight into how the interaction of person, situation, and strategy impacts well-being. Thus, the aims of this dissertation are to (1) assess the naturalistic use of emotion regulation as a function of context and person, one interaction at a time; (2) develop and test a dynamic training paradigm which adaptively pairs situations and strategies to be implemented in real world contexts; and (3) assess the longitudinal effects of the implementation intentions training paradigm and how it is impacted by individual cultural values
What 'Home' Means to Residents in the Houston Area
Get PDFThe idea of “home” is a multidimensional concept that encapsulates a variety of meanings, ranging from psychological orientations and physical locations to the relationships that unfold within. In the summer of 2024, members of the Greater Houston Community Panel were surveyed and asked, “What do you think of when you hear the word ‘home’?” and given an open space to write 1-2 sentences to describe what came to mind. This snapshot explores the ways in which area residents conceptualize home. In short, residents organized their thinking into four broad categories: 1) psychological orientations/attachments, 2) geographical location/built environment, 3) social relationships, and 4) activities. Additionally, while most residents have positive associations with home, some do not feel “at home” and associate the word with negative experiences, highlighting the fluid nature of home in the context of high stress
Creation of an Integrated Gas Sensor Platform using 2D Materials
No full textTwo-dimensional (2D) materials have shown promising gas sensing capabilities due to their high surface-to-volume ratio and excellent electronic characteristics. However, the weak signals generated by 2D sensors are typically measured using ultra-precise but bulky electrical characterization equipment. Despite their potential, 2D material-based gas sensors have not yet been widely integrated into Internet of Things (IoT) systems for real-time, wireless, and continuous gas monitoring in both urban and industrial settings. This work presents an integrated 2D material-based gas sensing platform that amplifies low-intensity sensing signals on-site and employs Bluetooth 5 technology for long-range, real-time data transmission. It facilitates high-quality data collection from remote 2D gas sensors, addressing modern needs for life safety, smart living, efficient production, and environmental preservation. The thesis introduces 2D material-based gas sensors, details the development of the wireless transmission system, and describes the fabrication processes for 2D sensors and antennas. The creation of this platform lays the foundation for artificial intelligence-assisted, data-driven gas sensing with 2D materials, offering a promising approach to overcoming the selectivity limitations of chemiresistive-type 2D gas sensors, which yields richer insights into sensing behavior and ultimately deepens the understanding of the dynamics of modern human living conditions