University of Maryland, Baltimore County
Not a member yet
17643 research outputs found
Sort by
DISCOVERY OF AURORA-A KINASE PHYSIOLOGICAL SUBSTRATES IN PROSTATE CANCER CELLS
No full textAurora-A is a serine/threonine kinase that has oncogenic properties in vivo. The expression and kinase activity of Aurora-A are up-regulated in multiple malignancies including prostate cancer, which is the most frequently diagnosed malignancy in men and the second leading cause of cancer death in the United States. Aurora-A is a key regulator of mitosis that localizes to the centrosome from the G2 phase of the cell cycle through mitotic exit and regulates mitotic spindle formation as well as centrosome separation. Overexpression of Aurora-A in multiple malignancies has been linked to higher tumor grade and poor prognosis through mechanisms that remain to be defined. Using an unbiased proteomics approach, we identified the protein Nuclear Mitotic Apparatus (NuMA) as a robust substrate of Aurora-A kinase. Using a small molecule Aurora-A inhibitor in conjunction with a Reverse In-gel Kinase Assay, we demonstrate that NuMA becomes hypo-phosphorylated in vivo upon Aurora-A inhibition. Using an alanine substitution strategy, we identified multiple Aurora-A phospho-acceptor sites in the C-terminal tail of NuMA. Functional analyses demonstrate that mutation of three of these Aurora A phospho-acceptor sites significantly diminishes cell proliferation. In addition, alanine mutation at these sites significantly increases the rate of apoptosis. Using confocal immunofluorescence microscopy, we show that the NuMA T1804A mutant mis-localizes to the cytoplasm in interphase nuclei in a punctate pattern. The identification of Aurora-A phosphorylation sites in NuMA that play a role in cell cycle progression and apoptosis provides new insights into Aurora-A function
Efficient Gossip Computations in Wireless Sensor Networks
No full textWe consider the implementation issues that arise in gossip-based distributed average consensus (DAC) algorithms for wireless sensor networks (WSN) with nodes operating at low duty cycles(1-5%). Gossip protocols provide a robust mechanism to handle the unreliability in communication arising due to the dynamic nature of WSNs. We present a practical implementation with improvements over simple gossip protocol that achieves DAC in low duty cycle WSNs. Our implementation deals with problems due to deployment, neighbor discovery, scheduling, convergence detection, and adaptability in a dynamic and changing WSN. Our implementation uses the Castalia simulation framework but it can be adapted to any other simulation framework or other WSN platform. We show that the proposed implementation techniques effectively solve these problems and demonstrate significant performance improvements for practical distributed consensus
Visualization of Smoke and Fire Data Based on the Fire Dynamics Simulation Model
No full textThis thesis presents a technique to visualize smoke and fire data generated from the NIST Fire Dynamics Simulation model using Monte-Carlo single and multiple scattering. I use established physics to extract the necessary data to visualize the smoke. The major challenge in smoke rendering with the fire simulation data is that the simulation resolution is not high enough for the visual effects of radiation occlusion. Standard volume lighting methods do not work well for fire since extensive areas can radiate as light sources, making the rendering prohibitively expensive. The proposed concepts of voxel lighting and spherical lighting mimic blackbody radiation in a unit area, approximating area diffuse lights with a sampling of point lights. I address this computational challenge with a selective light sampling scheme based on distance for fire smoke, and additionally with importance sampling of multiple scattering directions. This thesis also provides some analysis of a multiple scattering scheme based on Metropolis-Hasting sampling
Nimbus: Scalable, Distributed, In-Memory Data Storage
No full textThe Apache Hadoop project provides a framework for reliable, scalable, distributed computing. The storage layer of Hadoop, called the Hadoop Distributed File System (HDFS), is an append-only distributed le system designed for commodity hardware. The append-only nature of the le system limits the ability for applications to have random reads and writes of data. This was addressed by Apache HBase and Apache Accumulo, which both allow for quick random access to a highly scalable key/value store. However, these projects still require data to be read from the local disk of the server, and therefore cannot handle the type of I/O throughput that many applications require. This limits the potential for \hot data sets that cannot be stored in memory of one machine, but do not need the scalability of HBase, i.e. the ones that can be sharded and stored in memory on dozens of machines. These data sets are often referenced by many applications and be several gigabytes in size. Nimbus is a project designed for Hadoop to expose distributed in-memory data structures, backed by the reliability of HDFS. By executing a series of I/O benchmarks against HBase and fully integrated with MapReduce input and output formats. The following discusses relevant use cases and demonstrates Nimbus's performance advantage over HBase, allowing for high-throughput data fetch operations for performant applications
The Effect of Program Generosity on Variation in Per Enrollee Medicaid Spending by State
No full textIn this dissertation, the impact of program generosity on per enrollee Medicaid spending by state is examined through a combination of descriptive and econometric analyses. The dissertation employs a unique dataset that compiles a variety of Medicaid program data, including per enrollee spending and Medicaid program characteristics, covering the period 2002-2004. Generosity indicators are constructed to quantitatively measure differences in program generosity across three dimensions: optional benefits, eligibility rules, and physician fees. These indicators are the variables of interest in the econometric analysis of per enrollee Medicaid spending, while additionally controlling for relevant economic, demographic, health supply, ideology, and policy variables. Analysis of the variation in per enrollee Medicaid spending by service and by Basis of Eligibility (BOE), measured by the coefficient of variation, demonstrated that the magnitude of variation in each service or BOE category was inversely related to the magnitude of per enrollee spending. In other words, the services and eligibility groups with the lowest per enrollee spending demonstrate the greatest level of variation. Overall variation most closely resembles that of the categories with the highest spending. As a result, any effort to reduce variation would have to focus on those service or eligibility groups where per enrollee spending is highest. No one state is most or least generous across indicators, nor is generosity (or lack thereof) an absolute predictor of per enrollee spending. The generosity indicators, plus year dummies, control for about 13 percent of variation in per enrollee Medicaid spending. The complete spending model controls for 63 percent of the variation in per enrollee Medicaid spending. Economic and ideology variables control for the largest share of the variation in per enrollee Medicaid spending
Head Start Preschool Teacher Implementation Of The Goals Curriculum: A Classroom-Based Promotion Of Friendships, Self-Regulation, And Positive Social Expectations
No full textThe current study includes the second randomized trial of the GOALS program, a universal preschool social emotional learning (SEL) prevention program and examines the impact of the program on children's social and emotional development. The study includes three separate aims including (1) creating a training and support model for teachers, (2) examining child outcomes when teachers implement the GOALS lessons, and (3) examining the moderating effect of gender, age, and verbal ability on child outcomes. For the first aim, the primary investigator created a training and support model, provided teachers with detailed lesson scripts, and provided ongoing consultation. To examine fidelity, teachers rated themselves on fidelity characteristics (e.g., adherence, quality, child responsiveness). Trained observers also observed approximately half of the lessons to examine teacher adherence to lesson content. Overall, teachers rated the training and consultation sessions as helpful and supportive and reported an overall positive attitude towards the GOALS program. The large discrepancy between observer and teacher report of adherence suggests that teachers provide inaccurate ratings of fidelity. For the second and third aims, trained undergraduate students conducted baseline and post-test assessments. Baseline and post-test assessments included both teacher ratings of child behavior and direct child assessments conducted by trained undergraduate students. Results revealed that children randomly assigned to the GOALS classrooms improved in teacher-reported self-control and provided more socially competent responses compared to children in the standard practices classrooms. Moderator analyses revealed that for boys GOALS decreased attention problems, and for older children and children with more advanced verbal skills GOALS improved social skills. Results from the current randomized trial parallel results from the initial trial and suggest that GOALS is effective in improving positive aspects of social behavior and information processing
An examination of social and role functioning among Baltimore youth at risk for psychosis
No full textSocial and role functioning have been determined to be risk factors and an outcome of interest for individuals who develop a psychotic disorder. Less is known about the nature of functioning for individuals at-risk for a psychotic disorder. The current study examined the contribution of psychosis-risk symptoms to social and role functioning using multiple measures of functioning in a treatment-seeking sample of adolescents and young adults. Results of multiple regression analyses indicate that psychosis-risk symptoms contribute differentially to functioning based on the type of functioning (social and role), as well as the measure of functioning. Using an increased understanding of risk and general psychopathology to predict functioning within psychosis-risk samples may also provide valuable information that could aid early identification efforts to improve long-term functional outcomes, delay, or prevention of psychosis
Activating Anti-tumor Immunity by Overcoming Programmed Death Ligand 1-mediated Immunosuppression
No full textImmunotherapy is a promising approach for the treatment of primary and metastatic cancer because of the immune system's ability to recognize and target tumor cells. Tumor cells employ various mechanisms to escape anti-tumor immunity. One such mechanism they employ is mediated by programmed death ligand 1 (PDL1), which interacts with T cell-expressed PD1 and results in T cell apoptosis. PDL1 is expressed constitutively or is induced by interferon-gamma (IFNγ) on the surface of most tumor cells and protects tumor cells from T cell-mediated lysis. PDL1 also suppresses T cell activation by binding T cell-expressed CD80. Given that CD80 and PDL1 bind each other, work in this thesis will exploit this interaction to design a therapy that prevents PDL1-mediatied immunosuppression to allow tumor recognition by tumor-specific T cells. Therefore, it was hypothesized that overexpressing CD80 on PDL1+ tumor cells will functionally inactivate PDL1, and secondly that a soluble form of CD80 (CD80-Fc) would similarly inactivate PDL1. Initial experiments demonstrated that co-expression of CD80 by PDL1+ mouse and human tumor cells prevented the binding of PD1, and co-culture of CD80+PDL1+ tumor cells with PD1+ T cells maintained T cells in an activated state. Further work demonstrated that CD80-Fc restores activation of CD4+ and CD8+ T cells similar to that of membrane-bound CD80. Since monoclonal antibodies to PD1 and PDL1 are currently in clinical trials, the ability of CD80-Fc to prevent PDL1-mediated suppression was compared to that of anti-PDL1 and anti-PD1 antibodies. These experiments demonstrated that CD80-Fc is more effective at facilitating T cell activation than treatment with mAb to either PD1 or PDL1. Although CD80 is well known as a costimulatory molecule, using antibodies to block CD28-CD80 costimulation on T cells does not prevent CD80-mediated restoration of T cell activation, indicating an alternate role for CD80. In the absence of costimulation, CD80-Fc restores T cell activation, confirming that soluble CD80 mediates this effect by simultaneously neutralizing PD1-PDL1-mediated immune suppression and providing CD28-CD80 costimulation. These studies identify a new role for CD80 and suggest that it may be a better therapeutic agent than antibody therapy for overcoming PD1-PDL1-mediated immune suppression in cancer patients
The Haar Wavelet Motif Relating Decision Trees, Neural Networks, and Rule-Exception Sets
No full textThis work presents a three-fold adaptation of the Haar Discrete Wavelet Transform (DWT), demonstrating its modification to efficiently transform a multiclass- (rather than numerical-) valued function over a multidimensional (rather than low dimensional) domain, or transform a multiclass-valued decision tree into another useful representation. It is proven that this multidimensional, multiclass DWT uses dynamic programming to minimize (within its framework) the number of nontrivial wavelet coefficients needed to summarize a training set or decision tree. It is a spatially-localized algorithm that takes linear time in the number of training samples, after a sort. Convergence of the DWT to benchmark training sets seems to degrade with rising dimension in this test of high dimensional wavelets, which have been seen as difficult to implement. This multiclass multidimensional DWT has tightly coupled applications from learning dyadic decision trees directly from training data, rebalancing or converting pre-existing decision trees to fixed depth boolean or threshold neural networks (in effect parallelizing the evaluation of the trees), or learning rule-exception sets represented as a new form of tree called an E-tree, which could greatly help interpretation/visualization of a dataset
Learning Classifiers from Simulated Satellite Data and Applications to Environmental Monitoring
No full textRemote Sensing applications have gained a lot of attention in recent years due to their cost effectiveness over applications involving in-situ observations. We conduct experiments using the Community Radiative Transfer Model(CRTM) a widely used library for satellite observations to observe and detect high temperature points and concentrations of aerosols such as PM2.5 for a given region. The CRTM is a simulation library which requires atmosphere and surface data in the form of gridded profiles and solves the radiative transfer problem in clear or scattered atmosphere. The observations/radiances of CRTM assume microwave and infrared sensor. The observations/radiances by the sensor are in the form of channel brightness temperatures or radiances at multiple wavelengths. We simulate the atmosphere and surface conditions using the real whether data. The CRTM performs a forward operation and calculates a vector of channel brightness temperature for the given atmophere and surface conditions. We use the channel brightness data to build models to detect high temperature points and estimate the concentration of PM2.5 using classification and regression trees respectively. We test the models for different time periods and regions. The aim of the research is to create an observation operator using the CRTM which can be used where observations of satellite are required to evalute predictions made by some other simulation models