22689 research outputs found
Sort by
An Analysis of Modern Approaches to Measurably Define and Justify Resilience-Specific Energy System Improvements
Get PDFEnergy resilience is widely accepted as a prudent and necessary goal for a modernized electric grid. In the last 100 years, electricity has become critical to society. Being without power for a few hours can be costly, being without power for an extended duration results in extreme degradation of safety and security. A primary challenge in designing a resilient grid of the future lies in defining resilience as a quantifiable goal, not just a concept. This paper evaluates previous frameworks, attempts, and ideas to develop metrics or otherwise quantify resilience benefits as these enable investment in resilient-specific design. An evaluation of existing approaches resulted in developing a guiding algorithm to account for the challenges facing decision makers in assessing the actual value of a specific resilience improvement. A generalized list of concepts and means for quantification is also developed to validate the approach through practical discussion of data available to support claims. By establishing a balancing equation for quantifying the costs associated with resilience initiatives compared to the value of losses resulting from an unmitigated system, designs can be evaluated objectively for their probable, realistic resilience value
SINGLE MOLECULE FLUORESCENCE SPECTROSCOPY FOR QUANTITATIVE ANALYSIS OF DNA, RNA, AND NANOPARTICLES
No full textAs the primary protagonists of the central dogma of molecular biology, nucleic acids (i.e., DNA and RNA) carry genetic information within a biological system. The rich reservoir of biological information promoted a wide range of applications in clinical diagnostics[1] and therapeutics[2], [3], and the development of more advanced technologies with higher spatial and temporal resolution, sensitivity, quantitative capability, and throughput[4], [5]. The advancement of technology also in return opens new horizons in analytical chemistry, biology, and medicine. Single-molecule detection (SMD)[4] and microfluidics[6] are among these emerging technologies, which provide highly sensitive detection of nucleic acids and nanoparticles. Single molecule methods allow the observation[7] and manipulation[8] of the behavior of individual molecule at microscopic dimensions, which reveals the information masked by the ensemble-average techniques and allow for insights into the samples with complex composition. Microfluidic devices can be designed to precisely control and manipulate small volumes of fluids, which is essential for many single-molecule studies. Despite the advantages offered by single-molecule and microfluidic technologies in analytical applications, a critical limitation lies in the inefficient sampling of rare target molecules, hampering their sensitivity. Furthermore, there is a notable gap in methodologies providing multi-parametric and quantitative information for the analysis of complex systems with high heterogeneity.
This thesis addresses these challenges by developing a versatile single-molecule and single-particle analysis platform for DNA, RNA, and nanoparticles, offering rich quantitative information. The introduction of an in-line preconcentration technique, high-salt molecular rheotaxis (HiSMRT), enhances the analytical sensitivity of microfluidic-based single-molecule detection for nucleic acids. This electrode-free approach allows for DNA concentration and recovery under physiologically relevant ionic conditions (Chapter 2). Subsequently, quantitative analysis of RNA, and the exploration of RNA biophysical characteristics in different states are achieved through the single molecule free solution hydrodynamic separation (SML-FSHS) platform (Chapter 3). This platform lays the foundation for investigating nanoparticles encapsulating DNA and RNA. Improvements to the existing CICS system, transitioning to a three-color configuration, incorporating fluorescence coincidence analysis, and implementing a deconvolution algorithm, enable quantitative analysis of lipid nanoparticles (LNP) (Chapter 4). The thesis concludes by implementing two modes of single molecule detection for LNP analysis, integrating continuous flow and SML-FSHS modes. This comprehensive approach offers correlations of size, payload, and lipid contents, providing insights into the kinetic mechanisms of nanoparticle assembly (Chapter 5). Overall, the developed platform addresses the existing gaps, offering a powerful tool for the detailed analysis of complex biological systems at the single-molecule and single-particle levels
FINDING MYSELF THROUGH WRITING: A JOURNEY OF SELF-DISCOVERY
Get PDFI started out as a biologist; when I proved to be bad at lab work, I turned to the written word and enrolled in this program. This thesis is something of a journey of self-discovery, presenting increasingly personal pieces from my time in the program
Cortical Population Dynamics Underlying Choice, Reaction Time, and Confidence
Get PDFDecision-making is a fundamental aspect of behavior that humans perform ubiquitously and repeatedly throughout their lives. Some decisions are second nature, like deciding to attend class, while others are complex, requiring an abundance of information to consider, such as the decision of which university to attend. However, all decision-making processes share core attributes that encompass the spectrum of complexity. Over the last several decades, scientists have identified and probed these characteristics to unveil the computational and neural underpinnings that make decision-making possible. In addition, these advances have pushed our understanding in other cognitive domains, including perception, attention, memory, learning, reward processing, and others. The essential nature of decision-making has provided an avenue for investigating and comprehending complex cognitive functions at a mechanistic level.
The research laid out in this dissertation addresses the neural representation of evidence for a decision, its temporal dynamics and spatial distribution across two levels of the cortical hierarchy, and how it predicts three key behavioral manifestations: choice, reaction time, and confidence. Chapter One presents a general introduction of the perceptual decision-making field both from a behavioral and neurophysiological perspective. Chapter Two describes the novel behavioral paradigm I developed and the main two models I considered throughout this work. Chapter Three summarizes my findings in visual areas MT/MST and neural correlates of choice and confidence. In Chapter Four I detail my electrophysiological results from area LIP, suggesting a parallel strategy for simultaneous reporting of choice and confidence. Lastly, Chapter Five provides a systematic analysis of the training process used to prepare animals for the experiments, elucidating interesting possible theories for learning. All in all, this work demonstrates both the strengths and limitations of current theories within the framework of bounded evidence accumulation, bringing us closer to a comprehensive account of decision formation and confidence judgments
Toward Development of Novel Antimicrobial Strategies For Treatment of Bacterial and Viral Infection
No full textMulti-drug resistant infections continue to pose a significant threat to human health, highlighting an urgent need for novel antibacterial strategies. Central metabolic targets, with a focus towards narrow spectrum antibiotics, is one area that can be further explored for drug development. 1-Deoxy-D-xylulose 5-phosphate synthase (DXPS) is an exciting antimicrobial candidate as it sits at a branchpoint in central metabolism, is essential to bacteria, is absent in humans, and is distinct from related mammalian enzymes. The work in this thesis focuses, in part, on the alkylacetylphosphonate (alkylAP) class of antimicrobials that selectively inhibit DXPS. Chapter 2 explores their selectivity between two enzyme homologs of DXPS, the well-studied Escherichia coli (EcDXPS) and the recently cloned Pseudomonas aeruginosa (PaDXPS). Our findings demonstrate that the bisubstrate-analog alkylAPs, which are submicromolar inhibitors of EcDXPS are significantly less potent on PaDXPS, opening the door towards narrow-spectrum drug design. In Chapter 3, a siderophore-based strategy is explored for active uptake of alkylAPs into pathogenic bacteria. Three alkylAP-siderophore conjugates were designed that can either directly inhibit DXPS or be enzymatically hydrolyzed to release the inhibitor. This preliminary analysis offers an additional angle to narrow spectrum antibiotic design, through targeting uptake machinery expressed in uropathogenic Enterobacteriaceae.
The latter third of this work shifts focus towards long-acting antiretroviral (LA-ARV) formulation for treatment of HIV. Nucleoside reverse transcriptase inhibitors, such as emtricitabine (FTC), are currently incompatible with LA-ARV due to their inherent hydrophilic properties. In Chapter 4, emtricitabine (FTC) prodrugs are synthesized with the goal of incorporating them into semi-solid prodrug nanoparticles (SSPNs) or polymers of prodrug (POP). These FTC prodrugs were then evaluated in commercially available human compartments for their rate of FTC release. We discovered flexibility at the 5’-hydroxyl position that allows variable release of FTC while maintaining compatibility with SSPN formulation, rapid release of hydroxyl-linked FTC prodrug dimers, and exposed a potential issue with metabolism of FTC prodrug dimers linked by the exocyclic amine. Taken together, these studies continue to advance the suitability of carbamate/carbonate or carbamate/ester prodrugs of FTC for LA-ARV formulations, for the ultimate goal of complete LA-ARV regimens
Efficient Spatiotemporal Representation Learning Techniques for Videos and Remote Sensing Data
No full textSpatiotemporal representation learning refers to the process of extracting informative features from data that contain both spatial (related to space) and temporal (related to time) information. It is a fundamental problem in video understanding and remote sensing applications, as data often contain both spatial and temporal dimensions. For instance, video frames contain spatial details, and their sequence reflects temporal dynamics. Similarly, satellite images in remote sensing offer spatial information over time, revealing changes and patterns.
Given the time-consuming and expensive nature of acquiring labeled spatiotemporal data, this thesis initially focuses on exploiting abundant unlabeled data for representation learning. We propose three novel techniques that can learn robust representations from unlabeled data through self-supervised pre-training. The first technique introduces a novel adaptive masking technique for spatiotemporal learning with masked autoencoders, which learn representations by reconstructing masked tokens from visible tokens. The second method proposes Mixed Barlow Twins, an extension of the widely adopted Barlow Twins in self-supervised learning, by incorporating mixed samples interaction to reduce overfitting. The third method introduces a novel approach to learn representations from off-the-shelf remote sensing images through the denoising processes in diffusion probabilistic models.
Following pre-training, transfer learning is the common practice to adopt pre-trained representations to specific downstream tasks. Hence, in the second part of this thesis, we introduce novel unsupervised, semi-supervised, and fully-supervised techniques for video and remote sensing applications. Our contributions include attention prompt tuning for parameter efficient fine-tuning for action recognition, deep metric learning for unsupervised change detection, cross-consistency regularization for semi-supervised change direction, hierarchical transformer networks for change detection and pansharpening, spatial and interaction space graph reasoning for road detection, and deep image prior coupled with overcomplete networks for hyperspectral pansharpening. Through extensive experiments conducted on various video and multi-temporal remote sensing datasets, we demonstrate the effectiveness of our proposed pre-training and transfer learning approaches compared to state-of-the-art methods
STRATEGIC VULNERABLITIES OF US OFFSHORE WIND ASSETS: A “NEW” US BORDER REQUIRES A LONG-TERM SECURITY PLAN
Get PDFThe United States has set ambitious offshore wind power generation goals in support of its 2015 Paris Agreement commitments. However, climate change and the multi-polar geopolitical landscape will likely translate into significant vulnerabilities, particularly in the Atlantic, which is the focus of this research. Government and the private sector have spent the last twenty years addressing cybersecurity of critical energy infrastructure, but physical risks from extreme weather and/or sabotage have not been adequately considered. The Great Power Competition in the Arctic will bring near-peers close to US waters, and offshore wind distributed throughout the US economic exclusion zone will be attractive targets for hybrid warfare tactics. At present, the US has limited maritime capacity to protect those assets, and the regulatory risk assessment approach focuses on minimizing a wind project’s impacts on its surroundings and other activities. The US will need a national, long-term offshore wind security plan to address risks to offshore wind, and federal agencies will need to better incorporate future security into current research, policy, and deployment efforts. Historical case studies from the US Gulf of Mexico, Texas and Ukraine help better understand the baseline for energy assets exposed to extreme weather events and hybrid warfare, and points towards challenges OSW will face in the future. This effort than looks forward at potential OSW vulnerabilities, how a multi-stakeholder body might prioritize those potential vulnerabilities through multi-criteria screening, and the need for both proactive and reactive controls. The analysis also addresses some key government entities that will need to be heavily involved. A framework is then proposed for how a multi-stakeholder process can be conducted, including a gaps analysis for adequately protecting offshore energy assets, and followed by development of a National US OSW Strategy and Roadmap for Long-Term OSW Security. The paper concludes by providing sample recommendations for short-term exercises and data collection projects that will help inform the larger and longer gaps analysis and roadmap process
In-Vivo Neural Tissue Impedance Measurements with a 126 Channel Silicon Microelectrode
Get PDFApplication of electrical stimulation is an indispensable method within neuroscientific research for investigating neural circuits and connectivity. Optogenetic methods of activation and inhibition have dominated rodent animal model applications, thanks to a high degree of specificity and selectivity, but this success is not easy to implement in NHP (non-human primate) models. Further, electrical stimulation is a clinically accepted means of addressing a variety of human neurological conditions through modulating pathological neural activity with applied currents. Therefore, advancements made in delivering precise electrical stimulation in a research context are sooner to see translation towards human applications than optogenetics methods.
Currently, Neuropixels probes are demonstrated to be a highly effective intracortical recording platform with both high spatial resolution and channel counts. This entails a single device that can record several different neurons across multiple brain regions simultaneously. However, this high spatial resolution and programmable position selectivity has yet to be extended towards activating neurons, not just recording them. While much work pertaining to electrical neural stimulation predates Neuropixels, as do multichannel stimulating implementations such as the Utah array, these prior technologies operate on a scale orders of magnitude larger than the microscale distances that CMOS based devices like Neuropixels can resolve. Further still, all existing electrical stimulation systems operate with a large margin of uncertainty on the magnitude to which a delivered stimulus is attenuated throughout the chronic interactions between a foreign electrode and a local biological environment.
Here, investigation is undertaken to make use of prototype patterned microelectrode shanks with 126 uniquely addressable and permutable TiN electrode sites; with the objective to establish design requirements for a current steering device. Additionally, investigation is also performed to understand the differing magnitudes of impedance changes in-vivo, comparing outcomes between currents sunk to a traditional headscrew, or to various locations on the probe itself. Leveraging the customizable and dynamically specifiable nature of microelectrodes on this device, hereto referred to as a STIMAL probe, insights are expected to be gained on the failure modes of chronically implanted neural interfaces, and their local impedance contribution
The lateral habenula is required for maternal behavior in the mouse dam
Get PDFOne of evolution’s key demands is: survive long enough to reproduce. But mammals extend this single demand to also include parenthood, to improve the chances their own offspring also survive long enough to reproduce themselves. Parenthood is a long-term commitment requiring immense energy expenditures by the parent to see the young through to maturity. To maintain motivation for such a monumental task, parenting must be innately rewarding. How did evolution tap into reward circuits to motivate parenting? Previous work in the rat has implicated the lateral habenula (LHb), a conserved epithalamic nucleus, as a potential neuroanatomical intersection of parenting and reward circuitry.
In this dissertation, I examine the role of the LHb in maternal behavior in the naturally parturient mouse dam (mother). I show that kainic acid lesions produced a severe maternal neglect phenotype in the primiparous mouse dam towards her litter. Then I show chronic chemogenetic inactivation of LHb using inhibitory designer receptors exclusively activated by designer drugs (DREADDs) in mouse dams impaired maternal behavior in DREADD-treated dams compared to control- treated dams. Using a random intercepts linear mixed model for longitudinal maternal behavior, I conducted an in-depth comparison of group performance in pup retrieval and nest building. These two behaviors were chosen as examples of a novel component of maternal behavior, and an already-established maternal behavior, respectively. Finally, I examine spatial histology data from both sets of
experiments and examine evidence suggesting posterior LHb may house the ii
maternally-relevant cells, in an effort to provide a foothold for future work examining maternal behavior regulation in the LHb.
In the methods chapter, I describe an open-source home cage behavior data collection and analysis pipeline called Raspberry Pi Experiment (PiE), that I developed together with Dr. Robert Cudmore. Beginning with a physical 24” x 24” x 24” box, and then detailing the work conducted enabling any neuroscientist, regardless of coding background, to purchase a raspberry pi and wire their own home cage behavior box. With a user-configurable approach and the use of affordable and accessible supplies, an alternative to proprietary behavior boxes is detailed. Exemplar maternal behavior data is presented following data capture, scoring, and analysis using PiE
THE INTERPLAY OF HEALTH INFORMATION, HEALTH LITERACY AND CERVICAL CANCER SCREENING UPTAKE AMONG RURAL KENYAN WOMEN
No full textBackground: Most cervical cancer cases in Sub-Saharan Africa are diagnosed among rural women, yet rural women have low cervical cancer screening uptake rates. Lack of cervical cancer information is one of the major barriers to screening uptake in rural women. However, the impact of health information sources on cervical cancer screening uptake is rarely explored.
Objective: This research investigated the associations among health information, health literacy, and cervical cancer screening among rural Kenyan women with low educational attainment.
Methods: A convergent mixed methods study design was used. A convenience sample of women (N=174) completed a survey, and a purposive sub-sample (n=21) was interviewed in semi-structured interviews. Quantitative and qualitative data were analyzed separately, and results were merged on joint displays for comparison and interpretation.
Results: Every additional increase in the number of health information sources accessed was associated with an increase in health literacy after controlling for age and education (Beta coefficient= 0.16, p=0.03). Only 6.3% of women had ever been screened for cervical cancer. Every additional increase in the number of cervical cancer information sources was associated with 366% greater odds of ever being screened for cervical cancer (OR=4.66, CI=1.19-18.25). Most women (76%) were willing to self-collect samples for HPV-DNA testing. Having health workers (OR=1.88, CI=1.23-2.86) and the news media (radio and television) (OR=2.63, CI=1.27-5.48) as primary sources of health information, cervical cancer awareness (OR=1.39, CI= 1.50-8.11), and having heard of cervical cancer from the news media OR=2.43, CI=1.07-5.51), were associated with higher odds of self-sampling willingness. Needing to travel 30 to 120 minutes to the nearest health facility (OR=0.44, CI=0.20-0.93), and anticipated cervical cancer stigma (OR=0.71, CI=0.57-0.886), were significantly associated with decreased odds of self-sampling willingness. Qualitative and quantitative results mainly converged.
Conclusion: Increased access to health information was associated with increased health literacy and higher likelihood of cervical cancer screening uptake. Most women were willing to self-collect samples for HPV-DNA testing. Ensuring that self-sampling kits are available and accessible, provision of proper instructions on how to self-collect samples and demystification of cervical cancer stigma might enhance cervical cancer screening uptake among rural women