Environmental and Occupational Health Sciences Institute
Rutgers University Community RepositoryNot a member yet
58345 research outputs found
Sort by
SE(3)-equivariant convolutional neural networks for biomolecular interactions
No full textProtein-protein and protein-small molecule interactions are critical for biological function. A comprehensive solution to the biomolecular interactions problem could greatly enhance our understanding and development of biomolecules. Protein structural data plays an important role in creating models for these interactions, as it provides rich information for improving model generalizability and interpretation. This dissertation investigates deep learning approaches for biomolecular interactions through the use of SE(3)-equivariant convolutional neural networks (SE(3)-CNNs). These networks apply mathematical constraints to existing deep learning architectures to learn physically relevant, geometric features. SE(3)-CNNs provide a robust framework for addressing complex molecular biophysical challenges, leveraging representation learning to identify crucial interaction features and extend these insights to new, unseen data. In the first stage, we developed Dock2D to rapidly prototype solutions to the biomolecular interaction problem using deep learning. We create a synthetic 2D dataset containing receptor and ligand shape pairs, and prototype deep learning solutions based around two tasks, interaction pose (IP) prediction and interaction fact (IF). The first task is to solve docking two shapes together, the second is to determine if two shapes interact or not. We create the datasets with an underlying energy function parameterized by spatial transformations applied to a ligand shape relative to a receptor shape, where a lower energy is associated with stronger interaction. We implement fast Fourier transforms (FFT) for translational sampling and several rotational sampling methods to solve the two interaction problems, using fully-convolutional, SE(2)-equivariant convolutional neural networks (SE(2)-CNNs). These SE(2)-CNNs generate useful and generalizable interaction geometric features that reduce the need for rotational sampling and demonstrate that the IP and IF tasks are related through cross-evaluation. In the second stage, we apply the solutions to the IP and IF tasks to real data in 3D, we collect antibody-antigen complex structural data and their associated binding affinity values (), and expanded with homology model data augmentation. We implement SE(3)-CNNs similar in structure to Dock2D and construct a SO(3) angular grid adjacency network to extend the Dock2D solutions involving uniform rotational sampling orientation of the ligand with respect to the receptor. This model, SE3Dock, effectively solves the re-docking problem, defined as putting the known antibody-antigen complexes back together. Since we already know that the IP and IF tasks are related, we were able to joint-train a regression model for binding affinity prediction, however only achieved enough accuracy to rank binding affinity for unseen antibody-antigen complexes. The follow-up model, SE3Bind, is a reformulation that removes the need for spatial sampling, for both the IP and IF tasks based on the a new dataset generation procedure, that presents the receptor-ligand pairs centered and already in their docked complex. While this model requires further development and testing, it successfully solves the IP task. To extend the advantages of using the SE(3)-CNNs onto protein-small molecule interactions, we developed SE3Lig. This model does not use FFT-based correlation or energies, and instead predicts probabilities for 18 different atom channels by in-painting the atomic density of cofactors, water, and ions---onto 0.5~AA resolution grids. The in-painted probability densities can be numerically compared between predicted atomic channels or taken as composites to infer what molecule or molecular fragment binds, and how relatively likely within specific regions on a protein. The SE(3)-CNN allows for data efficient training, without data augmentation or rotational sampling for this task, resulting in rapid model training and evaluation. Since AI generated protein structures typically do not predict small molecule positions, our model is well suited for rapid screening of the plethora of AI predicted protein structures, for small molecule interaction.Ph.D.Includes bibliographical reference
Engendering college readiness in underserved and disadvantaged students: the role of early learning and early college programs in improving academic outcomes
No full textHistorical disparities in access to quality education among minority populations have contributed to significant achievement gaps in college readiness, college enrollment, and graduation rates. In response to these challenges, Early College schools were developed as part of a broader strategy to improve access to high quality education for at-risk and underserved communities. Although the successes of Early College schools are well-documented, there remains a scarcity of research that empirically investigates which programmatic resources, practices, and student contextual factors within these schools are most effective in building or, conversely, impeding college readiness for underserved students. This study aims to fill this gap by empirically examining how an Early College pipeline nurtures college readiness through the application of specific strategies, resources, and practices to build key cognitive strategies, academic behaviors, content knowledge, and college contextual skills. The LEAP Academy University Charter School in Camden serves as the case study for this research to provide a detailed exploration of how a comprehensive, cradle-to-college educational pipeline can effectively prepare underserved minority students for success in higher education. Utilizing the David Conley College Readiness Framework, this study investigates how various program components at LEAP Academy contribute to the development of the different aspects of college readiness as outlined in the framework. The research employs a two-phase sequential multi-method design, incorporating interviews with staff, students, and alumni, observations of students within the LEAP school environment, document analysis of historical LEAP data, and a survey to empirically identify the factors that influence college readiness at an Early College school.
The results suggest that building well-rounded students capable of succeeding in college requires the integration of multiple factors that collectively contribute to college readiness. These factors cannot be addressed in isolation; rather, they must work together to create a comprehensive support system for students. Specifically, eight interrelated practices and resources were identified as critical to fostering college readiness at LEAP Academy. The study also tested the association between a set of identified LEAP pipeline components and student contextual factors with college readiness, using a range of parametric and non-parametric analytical tools to estimate the impact of these factors on students' academic preparedness. The findings revealed that participation in advanced academic courses, such as Dual Enrollment, showed the strongest and most consistent association with college preparatory progress. Moreover, participation in academic enrichment activities and early childhood enrollment at LEAP were shown to positively impact the development of cognitive strategies, as measured by NWEA MAP indicators. This research reveals the need for more studies investigating the systems and structures utilized by Early College schools in addressing college readiness among underserved minority students.Ph.D.Includes bibliographical reference
Comparing sphingosine-phosphate receptor 3 expression in stress-related brain regions in males and females
No full textSphingosine-1-phosphate receptor 3 (S1PR3) is a G protein-coupled receptor that binds to the lipid signaling molecule sphingosine-1-phosphate (S1P). S1PR3 is found in various tissues and cell types throughout the body, including the central nervous system (CNS). While its exact role in anxiety is still being elucidated. Research suggests that S1PR3 may play a role in modulating anxiety-related behaviors and neurobiology through its actions in the CNS [3] S1PR3 is expressed in certain regions of the brain, including the amygdala, hippocampus, and prefrontal cortex, which engage in anxiety regulation. Activating S1PR3 in these brain regions may contribute to neurotransmission and neuronal activity associated with anxiety pathways Corbett et al., 2019. However, the role of S1PR3 in the hippocampus and mPFC in male and female mice and comparison during the estrus cycle has never been studied. Therefore, we proposed to compare the concentrations of S1PR3 in males and females during the non-proestrus and proestrus stages. We performed immunohistochemistry on the hippocampus and mPFC of the brain and collected the confocal images, then counted the cells on image J. The data we obtained was analyzed on R.M.S.Includes bibliographical reference
The development of China’s community care services and its impact on older adults' health and subjective wellbeing: 1998 -2018
No full textAs China has experienced rapid population aging, the demand for elder care services has increased dramatically. The gap in service demand has become concerning because the traditional provision of eldercare heavily relies on informal care provided by the family. Social and demographic transformations have made the family-reliant care model unsustainable. Community care has become a reasonable and practical option to balance the care demand. Despite the crucial role of community care in elder care provision, the focus of many scholars has predominantly been on state, market, or family-provided care, with less attention given to community-based services.This dissertation examines the evolution and impact of community care services on the health and subjective well-being (SWB) of older adults in China. Utilizing data from four waves of the Chinese Longitudinal Healthy Longevity Study, the study employs person fixed-effect regression analysis to assess community care influence on the broader population of older adults, including subpopulations such as the childless and disabled.
Key findings indicate a significant disparity between the provision and needs of community care, with unmet needs having a notable negative impact on both health and SWB,especially pronounced in urban areas. The research reveals that while community care services significantly enhance SWB and health, the impact varies depending on the type of service, place of residency, disability status, and financial sufficiency. Specifically, cultural and social services were found to have a more significant effect on health outcomes, whereas medical care services notably influenced SWB. Importantly, the study also found that community care services might more directly influence older adults' subjective well-being than their objective health outcomes.
The dissertation highlights the critical need for policies that foster an integrated care model combining community and family care to effectively support aging in place. This approach not only leverages the strengths of both care systems but also addresses the unique needs of the aging population, thereby enhancing their quality of life and well-being. The study advocates for future research to incorporate more objective health measures and to examine the influence of degree of urbanicity on the effectiveness of community care services.Ph.D.Includes bibliographical referencesIncludes vit
Value and choice across the orbitofrontal cortex
No full textAs we go through our lives, we are perpetually using information about our environments to construct value judgements. These judgments, in turn, inform how best to interact with the world around us. For primates, both human and nonhuman (NHP), the orbitofrontal cortex (OFC) is essential for guiding this type of value-based decision making behavior. However, the mechanisms through which OFC contributes to the processes underlying value-based decision making are not clear. Our current understanding of these mechanisms is limited by three critical factors. First, we lack a behavioral paradigm that adequately captures important features of real-world decision making. Second, much of what we know about how the OFC contributes to decision making has been gleaned from the study of individual neurons. However, the responses of neurons in the OFC show considerable heterogeneity, making the information carried by individual neurons poor reflections of the population as a whole. Finally, and perhaps most importantly, the primate OFC is comprised of several distinct subregions that differ in both their cytoarchitecture and their connections with other parts of the brain. Despite this, the OFC is often treated as a single functional unit, thereby ignoring possible functional specializations across these subregions. The goal of this dissertation is to address these three issues. Chapter 2 directly addresses the address the absence of a suitable behavioral paradigm by validating a novel decision making task for macaque monkeys, one that forgoes strict oculomotor control and allows subjects to evaluate decision offers using natural free viewing behaviors. Chapter 3 simultaneously addresses the latter two issues by leveraging high-channel-count electrode arrays to conduct population-level neurophysiological recordings in two distinct subregions of the OFC. Together, the work presented in this dissertation will advance our understanding how the OFC contributes to value-based decision making.Ph.D.Includes bibliographical reference
Local anesthetic systemic toxicity (LAST) and cardiac arrest: assessment of simulation-based education training for labor and delivery staff
No full textPurpose of Project: The purpose of this project was to utilize simulation training to provide labor and delivery RNs, anesthesia providers, and OB-GYN residents with the necessary skills and knowledge sets to quickly identify local anesthetic systemic toxicity (LAST) and implement prompt treatment and/or initiate ACLS protocol if the patient progresses to cardiac arrest. The project sought to develop communication and teamwork skills amongst labor and delivery RNs, anesthesia providers, and OB-GYN residents. MethodThis is a quality improvement project that took place one two separate simulation implementation days that occurred six weeks apart. Each implementation day involved running 2-4 simulation rooms based on patient census on the labor and delivery unit. Each room had two project champions facilitating and documenting the scenario. An educational debrief and presentation was provided following implementation day one. The exact same scenario was then used on day 2 six weeks later, followed by another debrief summarizing the experiences and reinforcing educational resources. ResultsThe statistical methods used for the collected data were the Chi Square and Mann-Whitney U tests. The Chi Square test assessed skills met versus not met in implementation day one versus implementation day two groups. The was statistical significance for the increase in the number of skills that were met on day two when compared to day one (p value 0.0423 less than alpha value 0.05). A Mann-Whitney U test was performed and revealed that there is no statistical difference (U value 211 compared to U critical value 151) between the groups for the time that it took to complete each individual action item on the checklist. Implications The results of this project have implications for education development within healthcare organizations as well as clinical practice improvement. Simulation-based training can be used in a variety of settings and amongst all types of healthcare providers. When it comes to low frequency, high risk events such as LAST, it is imperative that providers can call upon skills and thought processes they may never have had to use previously.D.N.P.Includes bibliographical reference
Program evaluation for blood culture collection program in the adult emergency department
No full textPurpose of Project: There has been an increase in blood culture contamination rates in the adult emergency room at a local community hospital in NJ. The current rate of 3.953% exceeds the hospital's benchmark goal of 1.0%. A program evaluation of the blood culture collection program was completed utilizing the CIPP (Context, Input, Process, and Product) model to identify barriers and recommendations to improve the program. Methodology: The study used the CIPP model to evaluate the hospital's current blood culture collection program over 3 months. The context evaluation included analyzing the hospital's current practice to an evidence-based protocol and the shift the rates mostly occurred. The American Society of Microbiology recommends cleaning the injection site with alcohol chlorhexidine prep before injection with an allowed dry time of 30 seconds. The input evaluation included checking the availability of blood culture bottles and supplies needed. The process evaluated the competency skills of blood culture collection by 10 Registered nurses (RN) and 10 Patient care technicians (PCT) in a simulation lab. The product evaluated the data and made recommendations on areas of improvement. Results: The collected data analysis determined that 60% of contaminated rates occurred on the day shift and 40% on the night shift and that 100% of blood culture collection supplies were available at the time of collection. Competency assessment identified that 50% of RNs and 40% of PCTs appropriately collected blood culture samples. The ED was using an updated procedure while the education department was inconsistent with current standards of practice. Implications for Practice: The findings support that a program evaluation utilizing the CIPP model successfully identified the root cause of blood culture contamination rates and made appropriate recommendations. The CIPP model identified evidence-practice gaps impacting the increased rates.
Keywords: blood culture contamination rates, adult emergency room, blood culture collection protocol, CIPP modeD.N.P.Includes bibliographical reference
Program evaluation of fall prevention program
No full textPurpose of the Project: This Quality Improvement Project is a Program Evaluation of the Fall Prevention Program. Using the Context, Input, Process, and Product (CIPP) methodology, this project evaluated the implemented fall prevention program to answer the proposed clinical question, “Will a program evaluation of a community-based hospital's current fall prevention program reveal opportunities and evidence for change in practice that will facilitate improvement in the quality of fall prevention policy?”. Methodology: Each component of the CIPP methodology was evaluated and data collection revolved around chart analysis of 37 deidentified patient falls to analyze and evaluate fall trends. An anonymous staff survey to evaluate the knowledge of staff regarding the current fall prevention program was included in the data collection. Results: Gaps in practice were noted primarily in the synthesis of the product and process evaluation. Implications for Practice: It is believed that the incorporation of the proposed implications will strengthen the program, enabling it to perform well and meet its goals.D.N.P.Includes bibliographical reference
A dive into theoretical heat flux of diffuse flow hydrothermal vent plumes and applications to extraterrestrial oceans
No full text“Diffuse flow” is a general term for slow-moving, low temperature (0.2 °C to ~100 °C) fluids discharged from small cracks around hydrothermal vent systems. Sea water is heated up by subsurface magma, inducing a vertical movement. Using the remotely operated vehicle JASON, the vertical velocities of the plume heads discharging at the seafloor and their temperatures were measured at Axial Seamount, ASHES vent field in the eastern Pacific between the years 2018 - 2019. This was done to compare in situ samples with Cabled Array Vent Imaging Sonar (COVIS) data to estimate a total heat flux from diffuse flow vents. Observed vertical velocities are determined by tracking plume head rise from video particle-tracking analysis, and heat fluxes were calculated using 1- and 2-meter- tall thermistor arrays. The maximum temperatures were utilized from each position on the arrays for six distinct plume sites, averaging from top of array to vent opening 3.06 °C - 13.77 °C for the 2019 sites and 3.38 °C - 8.68 °C for the 2018 sites. The height on the array corresponds to the distance of the thermistor from the vent's orifice, the closer the thermistor, the lower the height. This approach proved more precise than relying on the mean temperatures, given the variability introduced by the movement of the arrays and turbulence around plumes. The vertical velocities of each plume head were calculated by segmenting the video footage into frames of either 0.5 or 1 second increments and tracking particle trajectory throughout each frame. Geological reference points were employed to stabilize the background, mitigating movement induced by shifting video frames. The mean vertical velocity across the nine tracked plume heads is approximately 0.08 m/s. The integration of thermistor array datawith vertical velocities allows for estimating the total heat flux for the seawater above seven vent openings. Additionally, the investigation into diffuse flow vents offers valuable insights into postulating the plausible environmental conditions found on exoplanets hosting hydrothermal vent systems. A subsequent phase of this research will entail using velocity and temperature data gathered from ASHES vent field to infer potential conditions, including an idea of heat dispersion, around hypothetical hydrothermal vents beneath Enceladus’ icy crust. This component is essential for the exploration aimed at uncovering life on celestial bodies outside of our planet.M.S.Includes bibliographical reference