22689 research outputs found
Sort by
Mechanisms of Gene Expression Regulation in the Germline of C. Elegans
No full textGerm cells bear the unique task of passing on their genetic content to the next generation through the process of meiosis. Gene expression in the germline must be tightly regulated to establish and maintain an immortal cell fate. C. elegans employ a series of endogenous small RNA (endo-siRNA)-mediated pathways to coordinate both the expression and silencing of genes in the germline. Silencing of foreign genetic elements and other genes not intended for germline expression occurs via canonical RNAi mechanisms, however, the process by which gene licensing occurs by the CSR-1 Argonaute and its associated small RNAs has remained elusive.
Our work demonstrates that the mechanism of CSR-1 gene licensing significantly depends on its function downregulating the expression of the highly conserved chromatin condensing factor, morc-1, by post-transcriptionally cleaving its mRNA transcript. Tuning morc-1 expression is critical; overexpression of MORC-1 results in its ectopic spreading across thousands of genes, and subsequently, condensation of these loci. This overexpression and spread of MORC-1 causes reduced expression of thousands of genes and ultimately leads to sterility. In wild-type animals, CSR-1 prevents this cascade from occurring, thereby ensuring that germline genes are properly expressed to support germline development and maintenance.
We also show that CSR-1 promotes gene expression in a second way. By repressing expression of multiple piRNA biogenesis factors, CSR-1 prevents excess biogenesis of piRNAs. Overexpression of piRNAs results in mis-regulation of other classes of small RNAs, including the secondary endo-siRNAs, 22G-RNAs. Because the piRNAs are potent primary triggers of germline silencing, we hypothesize that this role for CSR-1 in repressing piRNA expression is essential in maintaining fertility and the integrity of the germline
Emergent Syntactic Behaviors and Mechanisms in Neural Language Models
Get PDFKnowledge of syntax—the structure of phrases and sentences—is necessary for robust generalization in natural language processing tasks. One reason for the success of large pre-trained language models (LMs) is that pre-training teaches LMs to rely on syntactic features, rather than incorrect surface features like word position or memorized n-grams. However, it is unclear how pre-training induces these preferences, or how LMs encode syntax: Is a certain model depth or parameterization required to encode syntactic knowledge? Does the style of the pre-training data influence how much data is required? Is syntax efficiently encoded in a specific model region, or redundantly encoded for different inputs and languages?
In this dissertation, I investigate the syntactic abilities that emerge in language models through exposure to natural language. I first evaluate the ability of multilingual language models to perform syntactic agreement in various languages. I then use syntactic transformations to assess whether language models learn to preferentially rely on syntactic explanations of inputs, rather than positional or lexical heuristics; I also use transformations to investigate which architectural and dataset features are important for efficiently teaching LMs to prefer syntactic features.
I also adapt and employ causal mediation analysis, a causal probing method that allows us to observe which neurons are responsible for observed behaviors. Using this method, I analyze which neurons are responsible for syntax-sensitive behavior, and to what extent these neurons are shared across different kinds of inputs (and across languages in multilingual LMs).
The final portion of this dissertation discusses whether interpretability studies like these can help us improve how we train and evaluate language models
ADVANCING THERAPEUTIC STRATEGIES IN UROGENITAL PATHOLOGIES: TISSUE ENGINEERING, DATA SCIENCE AND BIOINFORMATICS INVESTIGATIONS
No full textThe urogenital system of our body includes organs involved in storing and transmitting urine and reproductive fluids. These organs suffer continuous insults due to the toxicity of urine and the mechanical stress due to cyclic discharge. Pathologies associated with these organs have been hard to cure due to the cyclic nature of this physicochemical insult. One such tissue of interest in this work is urinary bladder. Tissue engineering techniques have been successfully demonstrated to support the regeneration of connective tissues by using biomaterials. Biomaterials have been extensively researched for their pro-regenerative and modular mechanical properties and their modifications have even further enhanced these properties. A subset of these biomaterials, derived from animal tissues, have been particularly interesting due to their pro-regenerative response and have been used as scaffolding material for incidences of skin burn, etc. However, due to their limited mechanical properties, they haven’t been successful with bladder tissue regeneration.
In this work, I have demonstrated a simple biomaterial modification strategy that confers the biomaterial with superior mechanical and enzymatic properties, making the biomaterial useful in reconstructive bladder surgeries, where a portion of the pathological tissue is removed and replaced with a similar patch of the biomaterial. The regenerated tissue was also analyzed for differences at the transcriptomic level using single-cell RNA sequencing (scRNA-seq). This analysis helped us pinpoint the differences at the pathway level between healthy and regenerated tissue.
The urethra is another organ of the urogenital system and it also suffers from a condition called urethral stricture, characterized by constriction of the urethra lumen, leading to difficulties in discharging urine. In this work, I have also analyzed the stricture tissue at the single-cell level to highlight the different subsets of cell populations observed in the tissue.
The single-cell analysis tools developed above were used to study the progression and therapeutics of Prostate Cancer. The publicly available single-cell datasets for prostate cancer were gathered together and integrated into a single dataset. The integrated dataset was used to identify the therapeutic potential of immunologically important gene targets, along with some other targets of interest in literature, including B7-H3. The integrated dataset helped us select only the tumor cells and compare their expression profile with the profile of healthy epithelial cells of the prostate. This, in turn, helped us hypothesize the effectiveness of various targets based on the expression levels and variation in those levels. The integrated dataset also serves as a resource to explore important questions related to Prostate Cancer in the future
Upgraded 90 GHz Detectors for the Cosmology Large Angular Scale Surveyor
No full textStudies of the cosmic microwave background (CMB) have revolutionized our understanding of the universe over the last half-century. Increasingly precise measurements of the temperature fluctuations of this primordial radiation have elucidated the evolution of the universe, yielding the ΛCDM standard model of cosmology. The last decade has brought new goals to the scientific forefront, as measurements shifted from the temperature fluctuations to the polarization of the CMB. The Cosmology Large Angular Scale Surveyor (CLASS) is a polarization-sensitive telescope array located at an altitude of 5,200 m in the Chilean Atacama Desert. The instrument is designed to measure "E-mode" (even parity) and "B-mode" (odd parity) polarization patterns in the CMB over large angular scales with the aim of improving our understanding of inflation, reionization, and neutrino masses. CLASS focal planes consist of arrays of highly sensitive feedhorn-coupled detectors incorporating transition-edge sensor (TES) bolometers. CLASS is currently observing with three telescopes covering four frequency bands: one at 40 GHz (Q); one at 90 GHz (W1); and one dichroic system at 150/220 GHz (G or "HF"). This thesis provides context for the cosmological research done by the CLASS experiment, and details novel instrumentation that was developed to study faint signals from the polarized CMB across large angular scales. I discuss the updated design, assembly, in-lab characterization, and on-sky performance of new 90 GHz detectors. Design changes were made to the TES bolometer architecture, with the aim of improving stability and optical efficiency. We assembled and tested four new detector wafers. These detectors were installed into the W1 telescope, and achieved first light in the austral winter of 2022. The new detectors had similar operating temperature, noise power, and saturation power to the originals. The median efficiency and the spread of efficiencies (68% interval) improved from 42%^+15_-22 to 60%^+10_-32. For the upgraded wafers alone, median efficiency increased to 65%^+0.06_-0.06. The overall noise-equivalent-temperature of the 90 GHz focal plane improved from 19 μK√s to 8.7 μK√s
MEASURING GENETIC COUNSELOR GOAL ADAPTATION AT THE INITIATION OF THE GENETIC COUNSELING SESSION
Get PDFAs demand for genetic counseling services continues to rise, there is an increasing need to develop models of practice to enable genetic counselors to practice at the top of their scope. To this end, the Reciprocal Engagement Model (REM) and its associated 17 genetic counseling goals have been developed to define genetic counselor practices and support the development of genetic counseling outcomes. The REM and its goals emphasize the need for genetic counselors to dynamically tailor their interactions to patient needs. However, the limited research that exists suggests genetic counselors may only marginally change their communication strategies to adapt to expressed patient needs.
In this cross-sectional mixed-methods study, genetic counselors responded to simulated referrals and patient vignettes, ranking REM goals and providing recorded responses to assess goal adaptation to patient distress. Participants also provided debriefing responses to generate context and elaborate strategies they utilized in reranking these goals.
Participation was open to currently practicing genetic counselors in the United States and Canada between June 1st to October 31st, 2023. 38 responses were included in the quantitative ranking analysis while 34 qualitative responses were analyzed. Counselors exhibited ranking differences in 5 of the REM goals with two goals having significant pairwise comparisons between specific arms of the study. Qualitative analysis of contracting responses showed more emotionally resonant language used by counselors to address the distressed patient, while debriefing responses provided additional context to reranking strategies and potential future directions for development of the REM. These results indicate genetic counselors recognize the need to adapt to stated patient needs while also suggesting further development opportunities for the REM
Fostering Graduate Student Creative Problem Solving in a Professional Military Education Context
Get PDFIn military contexts, a tension exists between the need for rapid, unquestioning obedience to orders, especially early in one’s career, and the need for senior leaders to solve complex problems creatively. For officers in the Marine Corps, a key milestone in their careers is the Marine Corps’ Command and Staff College, an intermediate-level professional military education master’s degree program. In 2015, the College, and the wider Marine Corps University community, established a plan to improve student creative problem solving; however, the plan did not meet its outcome goals by 2021.
The purpose of this study is twofold. First, using a convergent parallel mixed methods design, this study examined factors related to creative problem solving and their application to Command and Staff College curriculum. Key results of interviews, surveys, and secondary data analysis included the perceived need for additional time for students to think creatively, and the need to address the tension between authoritarian thinking and the imperative to develop new creative solutions. The second part of this study examined an intervention designed to give students more time to think and to give them structural, metacognitive supports for their thinking. Using a quasi-experimental design, the two key factors of concern for the study were metacognition and creative problem solving.
Improvements in the students’ metacognitive abilities were expected to lead to improvements in their creative problem-solving ability. Quantitative results showed no significant improvement in creative problem solving while there was actually a significant decrease in perceived metacognitive ability for both the comparison and intervention groups. According to explanatory interviews, one key factor in these results may have been the use of a perception survey, in which decreases in one’s perception of one’s metacognitive ability might mask actual improvements in real metacognitive ability. Another factor that emerged from the explanatory interviews was the need for the intervention to be more fully integrated across the whole curriculum. This study underscores the difficulty of making significant changes to student creative problem solving, especially in a military community. Further study could examine the relationship between perceptions of metacognitive ability and actual metacognitive ability
Decoding the transcriptional network that defines a specialized secretory organ
Get PDFThe biological process of secretion is a functional necessity for the survival of all organisms. The cellular mechanisms that mediate secretion require coordinated efforts between a wide range of genes involved in protein maintenance and trafficking. Higher organisms have developed distinct cell types and tissues that are uniquely specialized in secretion. While these organs play an essential role in organismal well-being, the early regulatory networks that define their capacity to function as a specialized secretory organ during development are less understood. The salivary gland (SG) of Drosophila melanogaster is an excellent model for studying the development and specification of secretory organs. The gland is a simple epithelial tubular organ, and much is known about the initial transcriptional processes that dictate SG cell fate. My thesis work focuses on the biological functions of the suite of transcription factors (TFs) that are activated and maintained throughout SG development in Drosophila melanogaster. Using single-cell RNA sequencing we captured a comprehensive transcriptome in the developing SG, which I then use to curate a list of genes that code for secreted products (termed “secretome”). I then adapted our single-cell pipeline to further characterize the distinct roles of CrebA and Rib – two TFs expressed during SG formation that were previously shown to boost secretory and translational capacity, respectively. Here I found that both TFs regulate operations beyond what we initially anticipated. We found that CrebA also plays a role in regulating translation while Rib regulates translation and energyproduction through mitochondrial processes. I then employed chromatin immunoprecipitation with sequencing to investigate the binding properties of Fkh and Sens- two additional TFs previously shown to regulate expression of the SG secretome in collaboration with a third TF (Sage). Here, I show that Fkh and Sens share a majority of their bound gene targets and that together they bind genes associated with a wide array of biological functions beyond the SG secretome. Lastly, I introduce a series of translational experiments that aim to probe the function of the SG exclusive TF Sage and its ortholog in Anopheles gambiae mosquitoes as a potential genetic tool in vector control
Application of SPSA-Type Algorithms in High-dimensional Stochastic Optimization and Sampling
Get PDFThis work mainly focuses on the two application scenarios of SPSA-type approximation. The first research topic is about the SPSA-type approximation in high-dimensional optimization problems, e.g. training of deep neural networks. The second is about the SPSA-type approximation in gradient-based sampling algorithms, e.g. Langevin Monte Carlo (LMC)
NOVEL APPROACHES FOR QUANTITATIVE PROTEOMICS
No full textThe accurate and precise identification and quantification of proteins and their modifications using mass spectrometry in large-scale studies play a central role in biological research. Despite significant advancements in mass spectrometry-based proteomics, where the identification of over 4000 proteins in a single run is now routine, the ability to accurately quantify proteins remains a persistent challenge. To address these issues, this work aims to identify the current problems in the field and provide effective solutions. The first part of this research involved evaluating different instrument platforms (Q Exactive, Exploris 480, and Lumos) and acquisition modes (DDA, PRM, and DIA) to determine which provided the most accurate, precise, and linear quantitative data. The results indicated that the Lumos instrument, utilizing DIA acquisition mode, demonstrated superior quantitative performance. The subsequent phase of this work addresses significant shortcomings in the analysis of large-scale proteomics experiments, particularly in terms of transition selection, protein quantification, and sample comparability. To overcome these challenges, a software tool named STraSE was developed. STraSE autonomously learns from the provided dataset and performs tasks such as transition selection, identification of quantifiable proteins, and ranking and filtering of samples based on data quality. To our knowledge, the STraSe software is the first to assess protein- and sample-specific quantification suitability. Lastly, this research focuses on the application of a method for the absolute quantification of proteins of interest. The method employs an internal calibration curve using four isotopologue calibrants to establish a linear response within each sample, allowing for accurate determination of sample amounts using interpolation. These strategies are vital in elucidating the stoichiometry of proteins and their post-translational modifications within complex mixtures for understanding biological mechanisms and disease. This knowledge can provide valuable insights into potential markers or factors contributing to temporal, environmental, or disease associated perturbations of healthy homeostasis
FRAGMENTATION PHENOMENA IN METALS UNDER HYPERVELOCITY IMPACT
No full textHypervelocity impacts represent a significant threat in low-earth orbit and hypersonic flight
applications, as well as an ever increasing performance demand for protection materials. The
analysis of such complex, heterogeneous mechanical events demands understanding of the full gamut of dynamic phenomena involved. Failure, fracture, and fragmentation are often the final stage of the extreme loading and deformation histories experienced by bodies under impact. Due in part to this terminal status and to a challenging combination of involved length scales, fragmentation is still one of the least tractable phenomena at play in extreme engineering applications. This works investigates three manifestations of fragmentation phenomena in hypervelocity impacts that span a wide range of timescales and energy densities.
First, we investigate the nature of the bright optical flash frequently accompanying highspeed ballistic impacts. This flash contains the signatures of the critical mechanisms operating at the earliest times after contact, the impacting materials, and the impact environment. Here we perform experiments investigating the physical structure and emission characteristics of the impact flash generated by 3 km/s spherical projectile impacts on structural metals using temporally co-registered high-resolution diagnostics. Experiments reveal that under these impact conditions and environment, the unexpectedly intense impact flash is the result of the rapid ablation of micron scale ejecta particles and fragments, accelerated to speeds in excess of 10 km/s by an impact jet originating from the contact point between projectile and target.
Next, we investigate the statistical fragmentation characteristics of rotary-swaged commercially pure aluminum powder compacts under non-perforating impact on hardened steel (AR500) anvils. We perform fragment soft-capture experiments to compare the statistical distributions of fragment size and shape of pure aluminum compacts of various processing. Results suggest that powder swaging may provide a method to produce metals that dynamically fragment more aggressively, with respect to distribution of fragment mass across size, than their quasi-static toughness would suggest.
Finally, using a new impact driven target configuration, we demonstrate the two-dimensional fragmentation of ductile metals under very-high strain-rate (10,000 1/s) equibiaxial tension and investigate the potential benefits of such a configuration for the lab-scale study of failure and fragmentation in ductile materials