Environmental and Occupational Health Sciences Institute
Rutgers University Community RepositoryNot a member yet
58345 research outputs found
Sort by
A model to predict the phenotype for copy number variants of uncertain significance
No full textPrenatal testing aims to identify disorders and birth defects, allowing for early interventions and better outcomes. Advances in genomic technologies like Next Generation Sequencing and Chromosomal Microarray Analysis have expanded the scope of these tests, going beyond detecting extra or missing chromosomes to identifying sub-chromosomal abnormalities and single-gene disorders. Copy number variations (CNVs), which involve duplications or deletions of large DNA segments, are a key source of genetic variation and are associated with complex disorders. These conditions involve multiple genes and can also be influenced by environmental factors. However, CNVs present challenges in interpretation due to the lack of objective criteria and uncertainty regarding their significance. Receiving Variants of Uncertain Significance (VUS) results presents clinicians with several challenges, such as whether to disclose these results to patients and how to counsel them. The results can also impact clinical management and necessitate follow-up studies. The issue of reclassification of a variant adds another layer of complexity. Described as placing patients in 'genetic purgatory', VUS results can be hard for patients to grasp, complicating their decision-making and potentially causing psychological distress.We discuss these challenges and suggest a pathway-based approach for better prediction of CNV variants in prenatal testing. In contrast to existing variant disease databases, which document pathogenicity without assessing likelihood, our approach is designed to provide a confidence score for pathogenic entries within dbVar. Our results illustrate the superior efficacy of employing a pathway-centric methodology in prenatal testing, facilitating more accurate prediction of CNV variants.Ph.D.Includes bibliographical reference
Nutritional factors affecting diet quality and endotoxin-induced inflammation
No full textDietary needs change through the life cycle and into older adulthood the ability to absorb and utilize nutrients declines. With aging, there is also a propensity toward weight gain and increased risk of inflammatory conditions such as diabetes, cardiovascular disease, and osteoporosis. Therefore, it is essential to determine optimal dietary patterns that provide adequate intake of beneficial nutrients while preventing excess consumption of other dietary components augmenting risk of adverse health outcomes. For instance, consuming adequate dietary protein is frequently met with the difficulty of limiting dietary fat. Dietary protein is essential for maintaining lean body mass (LBM) and bone in older adults; however, dietary protein sources contribute the largest portion of total and saturated fat in the American diet. Consuming excess saturated fat increases risk for dyslipidemia, obesity, cardiovascular disease, and metabolic syndrome. The balance of these macronutrients in an omnivorous diet, where a significant portion of both fat and protein come from meat, poultry, and dairy, is crucial to understanding benefits and risks in the older adult population. Therefore, the objectives of this dissertation were to evaluate benefits of dietary protein on body composition and diet patterns during weight loss, and to determine risks associated with dietary fat intake related to inflammatory components that activate innate immunity (i.e. endotoxin via NF-kB pathway).
The first study examined the influence of dietary protein on lean body mass and diet quality during weight loss in adults with overweight and obesity. This analysis included pooled data of multiple trials of 207 adults reporting food intake over 6 months of weight loss. The sample population was retrospectively divided by median protein intake (lower and higher protein). Our results showed that greater protein intake during calorie restriction attenuates loss of LBM and improves diet quality components by increasing intake of green vegetables and reducing intake of refined grains and added sugars.
The food study addressed method development for detecting and quantifying lipopolysaccharide (LPS) in foods. Fifty foods were assessed using the recombinant Factor C assay (rFC) and further examined in experiments assessing the effect of freezing and filtering food samples. Grains which are known to contain LPS-producing gram-negative bacteria were further assessed with the human Toll-like receptor 4 (TLR4) cell reporter assay. It was observed that LPS can be detected in food samples using both the rFC and hTLR4 cell reporter assays after filtration. Furthermore, the rFC assay is beneficial for detecting very low levels of LPS, whereas the hTLR4 cell reporter assay detects more inflammatory structures of LPS. Both assays provide a method of characterizing presence of inflammatory pathogen-associated molecular patterns (PAMPs) within food systems.
The last study in this dissertation is a trial using a randomized controlled crossover design examining how two levels of short-term dietary fat affected serum endotoxin and associated inflammatory markers in individuals with obesity compared to those with normal body weight. Thirty-two older adults were randomized to diet order of lower (20%) and higher (40%) fat diets. The group with obesity had a greater fasting serum LPS following HF diet compared to LF diet, whereas the group with normal body weight did not change. While HF diet increased serum LPS with obesity, the inflammatory markers increased in the entire sample. Our findings suggest that individuals with obesity are more susceptible to an elevation in serum LPS due to greater fat intake, but this does not consistently affect serum markers of inflammation.
Overall, it can be concluded that diet quality is altered by protein intake, and that dietary endotoxin and fat could contribute to diet-induced inflammation. The findings from this dissertation elucidate the intricate physiological responses to diet patterns that contain a combination of macronutrients (moderately high fat or protein) and PAMPs that influence health outcomes in older adults.Ph.D.Includes bibliographical reference
Examining the mechanisms of silver ion intoxication and detoxification in Staphylococcus aureus
No full textThe antimicrobial properties of silver (Ag) have been known and applied to fight infections for thousands of years. Silver shows promise as an antimicrobial because it can impact several cellular processes simultaneously. However, the mechanism by which the silver inactivates the cellular processes and how the cell combats this toxic effect is largely unknown. As the antibiotic resistance crisis continues to accelerate, human bacterial pathogens such as methicillin-resistant Staphylococcus aureus (MRSA) are becoming increasingly prevalent with rising mortality rates highlighting the need for additional antimicrobials. To effectively administer an antimicrobial, it is important to understand its mechanism of action. This study investigates how the cytosolic accumulation of silver ions alters the physiology of S. aureus. Specifically, we examine the pathogen’s response to Ag, examine how it detoxifies Ag, and demonstrate that Ag toxifies cells by altering carbon flux through the pentose phosphate pathway. Additionally, mutations identified from a suppressor screening show that inhibited transcriptional and translational functions promote survival in the presence of silver ions.M.S.Includes bibliographical reference
Phonological learning in the presence of lexical exceptions
No full textIn this dissertation, I establish a research program that uses computational modeling as a testbed for theories of phonological learning. This dissertation focuses on a fundamental question: how do children acquire sound patterns from noisy, real-world data, especially in the presence of lexical exceptions that defy regular patterns? For instance, Turkish infants tune into vowel harmony patterns as early as six months, despite lexical exceptions from disharmonic loanwords. This dissertation demonstrates that phonological learning is facilitated by two essential elements: (1) a restrictive hypothesis space defined by formal language theory and (2) an exception-filtering mechanism. I developed a learning model that harnesses the discrete nature of categorical grammars to filter out lexical exceptions based on statistical criteria adapted from probabilistic models. This hybrid model learns interpretable grammars that approximate acceptability judgments in behavioral experiments, demonstrating robust performance across various real-world corpora from English, Polish, and Turkish. Moreover, the dissertation integrates the proposed phonotactic model into learning morphophonological alternations. This approach is not only competitive on real-world corpora but also substantiated by experimental evidence.Ph.D.Includes bibliographical reference
A platform for duckweed production: improving and testing the duckweed production module (DPM®)
No full textDuckweed has the potential to be an environmentally and economically sustainable crop with commercial applications ranging from food and feed, wastewater treatment and environmental monitoring, biofuels and bioplastics, dietary supplements and pharmaceuticals, and integrated as a life support system for space exploration. Cultivation benefits of duckweed include clonal propagation, high growth rate, small plant size, and the ability to utilize the whole plant for most applications (no waste). However, efforts to cultivate duckweed in commercial settings have seen limited success due to the largely undomesticated nature of the plant. Abiotic and biotic parameters have not been defined and standardized, and optimal parameters may differ depending on the variety of duckweed cultivated. I hypothesize in this thesis that the benefits of duckweed cultivation are best realized in a controlled environment agriculture setting, using vertical towers to maximize biomass production. I show how system engineering improvements and growth media management, informed by testing and analysis of abiotic factors along with harvest yield measurements, may provide a path to successful commercialization of duckweed products and services. I also compare preliminary bacteria community profiling of two DPM® systems under disinfection (re-circulating aqueous ozone) and non-disinfection conditions to determine the extent of bacterial tissue colonization disruption due to ozone treatment (if any).
In Chapter 1, I discuss the engineering and abiotic improvements made over five years of trials. This included scaling the original Duckweed Production Module (DPM®) system from three to six trays and developing management protocols to increase harvest yield as well as increase the production duration per crop.
In Chapter 2, A culture-independent approach was used to generate bacterial community profiles from two DPMs, one with intermittent aqueous ozonation of re-circulating growth media, and one without aqueous ozone treatment. The community profiles were compared to test if aqueous ozone treatment would significantly alter the prevalence and relative abundance of bacterial taxa associated with duckweed tissue when compared to the untreated tissue. Colony Forming Unit (CFU) was also compared to test the efficiency of aqueous ozone to decrease total bacterial load.
In conclusion, this thesis provides evidence that duckweed can be cultivated long-term and harvest yield increased using a vertical multi-layered platform under controlled conditions when management protocols are developed and followed. Furthermore, this thesis profiles the bacterial community composition of duckweed grown in DPM® systems under two different conditions and describes the effects of aqueous ozonation on bacterial community diversity over time.M.S.Includes bibliographical reference
Fine-tuning large language models: from accuracy enhancement to bias mitigation
No full textAs the frontier of artificial intelligence (AI) continues to expand, Large Language Models (LLMs) have emerged as pivotal tools across a broad spectrum of applications. These sophisticated AI systems, powered by vast amounts of data and advanced algorithms, are revolutionizing industries. However, the efficacy and ethical implications of these models, particularly in terms of fairness, remain an area of active investigation. This research delves into the capabilities of Large Language Models for classification tasks, particularly emphasizing how prompt design influences model performance and fairness. Leveraging datasets from varied high-stake domains, we scrutinize the effect of biased and unbiased prompts on the inference outcomes of several cutting-edge LLMs, including GPT-3.5, GPT-4, Gemini-Pro, LLaMA2 (7b and 13b parameters). By evaluating group fairness and individual fairness metrics, our study aims to uncover the extent to which prompt bias can affect different models' decisions. Moreover, we fine-tune GPT-3.5 to explore its potential for enhanced performance and fairness in classification tasks. Moreover, we fine-tune GPT-3.5 to explore its potential for enhanced performance and fairness in classification tasks. While our findings confirm that prompt engineering and model customization can help enhance LLM performance in terms of accuracy, the implications for fairness are less definitive. This uncertainty underscores the complexity of disentangling and addressing the inherent biases within AI systems, highlighting the critical need for ongoing research and methodological innovation in this area. Our study contributes to the ethical AI discourse by underscoring the importance of deliberate prompt design and the cautious application of model fine-tuning, advocating for strategies that prioritize both efficiency and equity. Through this work, we aim to advance the conversation on responsible AI use, and the ethical deployment of LLMs.M.S.Includes bibliographical reference
Exploring the impact of online networks available to BIPOC applicants of genetic counseling master’s programs
No full textThe genetic counseling field remains largely homogenous with 89% of practicing genetic counselors identifying as Non-Hispanic White (National Society of Genetic Counselors, 2023). Although there have been many diversity and recruitment initiatives, these efforts have not resulted in a notable change in demographics (Channaoui et al., 2020). A number of online networks have emerged with the mission of helping prospective students, including: Minority Genetic Professionals Network (MGPN), Genetics Opportunities, Learning, Development, and Empowerment Network (GOLDEN), Genetic Counseling Prospective Student Network (GCPSN), and GC Chat on Discord. Given the initiative to recruit and sustain more diversity in the field, a look into the utilization of these networks by Black, indigenous, and people of color (BIPOC) applicants is an important next step. The purpose of this study is to determine the extent to which BIPOC applicants are using online networks and the perceived impact of these networks. We found that common reasons for joining these online networks include needing help with applications and seeking connection to other BIPOC individuals. These networks may be offering those with limited connections and resources a way to get the exposure and experience they need to have a more competitive application. Mentors and genetic counselors are found to be major sources of support in the application process and these networks offer an avenue of connection. Overall, the utilization of these networks increased feelings of connection to the genetic counseling field, with networks aimed to support BIPOC applicants seeming to deepen that connection.M.S.Includes bibliographical reference
In the thick of it: operationalizing the relationship between Black people, Black spaces, and Black political unity
No full textThis dissertation addresses the concurrent evidence that Black people are not a monolith and are frequently politically unified. To date, researchers have focused on the salience of racial group interests in the lives of Black Americans to explain this phenomenon. I challenge the prevailing wisdom about in-group racial attitudes and assert that Black political unity results from a Black person’s political learning within the Black community. To support my argument, I introduce a novel concept and survey measure called Black immersion, which captures how embeddedness in the culture and environment African Americans have developed in the United States since slavery socializes Black Americans across the Black diaspora and influences Black public opinion. Using cross-sectional and panel survey data spanning fifty years, I demonstrate how Black immersion is a dependable explanation for Black political attitude formation and Black political unity. The theoretical contribution of Black immersion is a shift away from the predominant explanation for Black political decision-making as a result of previously held in-group racial attitudes. Rather than assume Black Americans innately view the political world in a particular way, I assert and make central the social and environmental factors that inform Black political attitude formation. Methodologically, Black immersion advances Black political socialization research by enabling a nuanced approach to observing and understanding how various Black social networks and institutions simultaneously influence Black political behavior.Ph.D.Includes bibliographical reference
Causal collaborative filtering
No full textIn the era of information explosion, recommender systems have become crucial tools for fulfilling users' personalized and complex demands, finding widespread application in various real-world services such as e-commerce, job seeking, and social media. Collaborative filtering algorithms, among others, are fundamental algorithms that support the underlying mechanism of recommender systems. In general, collaborative filtering algorithms leverage similarities between users and items to provide recommendations. The intuition is that similar users should share similar tastes, and similar items may be liked by similar users. Traditional collaborative filtering methods, including memory-based methods such as user/item-based CF as well as learning-based methods such as matrix factorization and deep learning models, primarily focus on focus on mining or learning correlative patterns from data for matching. However, many real-world applications are driven by underlying causal mechanisms, and reliance solely on correlative learning can lead to practical issues such as Simpson's paradox, confounding bias, and echo chambers.
To address these challenges, this dissertation introduces a set of causal collaborative filtering methods based on the structural causal model (SCM) framework. These methods aim to advance from correlative learning to causal learning by formulating the essence of recommender systems as a "what if" question: what would happen if we recommend a certain item to a target user? This question is answered by causal preference using the -operator. By constructing causal graphs and applying causal inference techniques, these methods estimate causal preferences, offering a novel approach to collaborative filtering.
The dissertation presents causal collaborative filtering methods that address various issues in recommender systems through different types of causal graphs and inference techniques. It begins with mitigating Simpson's paradox using conditional intervention on directed acyclic graphs. Subsequently, it tackles unobserved confounders by incorporating unobserved variables into the causal graph and employing front-door adjustment. Finally, it addresses the challenge of echo chambers by modeling the dynamic and iterative process of recommender systems with loops in causal graphs and applying the back-door adjustment method. This comprehensive approach provides a significant advancement in the field of recommender systems by integrating causal learning into collaborative filtering methods.Ph.D.Includes bibliographical reference
Implementing next-generation sequencing as part of newborn screening in state public health laboratories: status and barriers
No full textNewborn screening (NBS) programs are vital public health initiatives, facilitating early diagnosis and treatment for thousands of infants annually in the United States. Originally targeting phenylketonuria (PKU) in the 1960s, NBS has expanded to encompass a broad spectrum of disorders, aided by technological advancements such as tandem mass spectrometry and next-generation sequencing (NGS). NGS, in particular, offers the potential to revolutionize NBS. While NGS presents numerous benefits, its integration into NBS faces various challenges, including technical feasibility, financial barriers, and logistical considerations. This study aimed to assess the barriers, expectations, and plans of state NBS programs regarding NGS integration, providing insights to guide advancements and inform stakeholders. It was found that most states are not currently using NGS technology though there is interest in implementing it in the future. The biggest barriers were lack of appropriate staff to support data analysis, bioinformatics, and genetic counselors as well as the cost of sequencing and volume of samples.M.S.Includes bibliographical reference