University of Massachusetts Boston
University of Massachusetts Boston: ScholarWorks at UMassNot a member yet
8618 research outputs found
Sort by
Identification of Aggressive Subtypes in Androgen Receptor-Dependent Castration Resistant Prostate Cancer Through Super-Enhancer Analysis for Lineages (SEAL)
Get PDFMetastatic castration-resistant prostate cancer (mCRPC) is a heterogeneous disease, characterized by diverse drivers of progression and mechanisms of therapeutic resistance. Super-Enhancers (SEs) are large clusters of enhancers that can robustly drive transcription. SEs are frequently found at key oncogenic driver genes and play a significant role in cancer occurrence and progression. This research aims to evaluate whether SE landscape can predict the aggressiveness of prostate cancer and distinguish unique subtypes of mCRPC. Using Super-Enhancer Analysis for Lineages (SEAL) on published high-throughput chromatin structure data, I examined SEs across samples representing various stages of prostate cancer progression. Compared to SEs in normal prostate and primary prostate cancer, mCRPC samples exhibited a distinct SE program. Notably, five distinct SE programs were identified within mCRPC: three linked to androgen receptor (AR)-dependent CRPC (designated AR-1, AR-2, and AR-3) and two associated with AR-independent neuroendocrine (NEPC) and double-negative (DNPC) subtypes. For each subtype, top SE-driven genes with elevated expression were identified. Within the AR-dependent subtypes, AR-1 and AR-2 were particularly aggressive with higher tumor growth, showing enrichment in epithelial-mesenchymal transition (EMT) and metabolic pathways, respectively. Importantly, TWIST1 and HNF1A were identified as top SE-driven transcriptional drivers in AR-1 and AR-2 SE subtypes, respectively. In public mCRPC patient datasets, distinct TWIST1+ and HNF1A+ tumor subsets emerged within the AR-dependent subgroup, with patients showing high HNF1A expression displaying poor clinical outcomes. Furthermore, HNF1A was involved in glycolysis process and tumor development, indicating its role in metabolic reprogramming of tumor cells. Overall, this study demonstrates that distinct SE landscapes in mCRPC are associated with tumor progression, therapy resistance, and lineage plasticity. The molecular classifications defined by SEs may guide therapeutic decisions and facilitate the identification of key oncogenic drivers within each CRPC subtype
Learning on the Go: Elements of a Mobile Learning Strategy
Get PDFAs a Learning and Development (L&D) professional working within the context of mobile learning, I spend my time addressing the issue of learning on the go. In this paper, I share the concept of a Mobile Learning Strategy canvas, created to assist a more strategic approach to mobile learning by L&D practitioners. I assert that the ongoing evolution of training and development, specifically within the realm of mobile learning, requires specific strategic frameworks to be effective. As work becomes more technical and more complicated, workers benefit more and more from training and development. I explore both strategy as a solitary concept and strategy as applied to either mobile (a mobile strategy) or learning (a learning strategy). Lastly, bringing strategy, mobile, and learning together, I propose a framework for a mobile learning strategy. By strategically placing the learner and their context at the center of a mobile learning solution, L&D professionals can better serve their learning audiences and better support their business
HACK24F: AI Conversations in Healthcare
Get PDFNursing students often complete clinical hours under the supervision of instructors in traditional hospital settings. However, obtaining individualized, consistent feedback from patients about their interactions with nursing students is often not feasible. This limits students\u27 ability to fully understand how their communication skills are perceived and how they can improve. Currently, there are no models that represent realistic real life conversations with patients. Most virtual simulation models used for nursing students provide scripted responses that do not feel genuine
HACK24F: The Empathy Architects
Get PDFProblem Statement: We want to help patients from underserved communities feel more safe, respected and comfortable in interacting with healthcare providers during the initial in-person encounters while also helping providers build empathy and implement socio-emotionally competent communication behaviors. Solution: Empathy Evolution is an AI-powered role-playing game that helps patients simulate their ideal communication with healthcare providers
Planning for the Future of the Easthampton Council on Aging
Get PDFThis report describes an assessment undertaken by the Center for Social & Demographic Research on Aging (CSDRA) within the Gerontology Institute at the University of Massachusetts Boston, on behalf of the ad hoc Senior Center Building Committee. The goals of this project were to document the features of other communities building processes, engage residents in preliminary conversation about their needs and preferences as they consider aging in Easthampton. Results of that work are described in this report and recommendations to the ad hoc building committee are provided. These recommendations are meant to provide an objective description of this preliminary community engagement and inform them as they put forth recommendations to the City as to how new space could meet some of the needs and preferences of a growing and changing population of older people in Easthampto
HACK24F: AI Audio Extractor
Get PDFI want to make a next.js website locally and then be able to hopefully deploy on Vercel. Within the website I want to be able to use AI to separate the instruments (vocals, piano, guitar, drums, bass, etc.) and also identify which notes are being played. I was thinking that we might be able to use an AI stem splitter to separate the audio tracks and use another AI model for note detection
The Provision of Healthcare for Older Adults in a Massachusetts Jail/House of Correction: Perspectives from Providers
Get PDFTHE PROVISION OF HEALTHCARE FOR OLDER ADULTS IN A MASSACHUSETTS JAIL/HOUSE OF CORRECTION: PERSPECTIVES FROM PROVIDERS November 2024 Laura C. Driscoll, B.S., Boston University M.S.P.T., Boston University DPT, MGH Institute of Health Professions M.S., University of Massachusetts Boston Ph.D., University of Massachusetts Boston Directed by Professor Jan Mutchler The aging population in U.S. jails is growing rapidly presenting significant challenges for healthcare provision within carceral settings. Older adults in these facilities often experience complex health conditions, including chronic diseases and age-related impairments, that require specialized care. Despite this, limited research has explored the experiences of healthcare providers tasked with caring for this vulnerable group in jails. This dissertation examines the perspectives of healthcare providers working with older adults in short-term confinement in a Massachusetts jail/house of correction, focusing on how the unique relational, structural, organizational, and policy-based constraints of the carceral environment shape their ability to deliver care. Drawing on a socio-ecological model of health promotion and Watson’s Theory of Human Caring, this study employs qualitative methods to investigate the lived experiences of healthcare professionals. Findings reveal that while providers strive to offer patient-centered and equitable care, they face numerous barriers, including limited resources, policy restrictions, and the inherent conflicts between custodial priorities and healthcare needs. The study also highlights the impact of COVID-19 on healthcare delivery in jails, exacerbating existing challenges. The results point to the need for policy reforms that prioritize geriatric care and improved healthcare models to support both the incarcerated population and the professionals serving them. These findings contribute to the growing body of literature on healthcare in correctional settings, with specific attention to the needs of older adults, and provide recommendations for improving car
Introducing A Practitioner\u27s Framework for Mobilizing Collective Action
Get PDFINTRODUCING A PRACTITIONER’S FRAMEWORK FOR MOBILIZING COLLECTIVE ACTION December, 2024 Matthew P. Bond, B.A. University of North Carolina Greensboro, M.S., University of Massachusetts Boston, M.P.A., Suffolk University, Ph.D., University of Massachusetts Boston Directed by Christian Weller Social dilemmas are situations in which individual and collective incentives are at odds, leading to suboptimal collective outcomes that make all individuals worse off. Modeling policy problems stuck in stasis as social dilemmas reveals new solution pathways that promise progressive change. The solution to a social dilemma is collective action, yet since the introduction of the collective action problem in the 1960’s, a solution to this problem that results in a repeatably effective process for mobilizing collective action has yet to be found. Research work on collective action has been mostly theoretical and, particularly regarding applied efforts to combat policy problems in field settings. This study introduces a framework that synthesizes the latest progress in collective action theory to resolve policy problems framed as social dilemmas by outputting issue-specific intervention strategies through a repeatable process. In order to test the framework, I selected a case of a policy problem stuck in stasis, used the framework to design an intervention strategy, and then tested my intervention strategy against alternatives to assess its effectiveness in resolving the problem. Inequitable hiring outcomes for senior positions across all major organizational and institutional settings in the U.S., particularly when focusing on disparities in minority representation in such positions, is an unresolved policy problem that can be modeled as a social dilemma. The National Football League (NFL) is an American institution that has a well-documented history of inequitable hiring outcomes for its head coaching body. Further, the NFL deployed an intervention strategy, the Rooney Rule, in 2003, yet this intervention has achieved no measurable effect on outcomes at all. I developed an intervention strategy from my framework and tested it against the Rooney Rule and the null in a live experimental setting with 661 participants who were asked to engage in simulated hiring and firing cycles over the lifespan of the Rooney Rule (approximately 25 “seasons”). The results of this experiment showed conclusive and substantial progressive effects of the experimental treatment derived from my framework, which significantly outperformed the Rooney Rule and the null while controlling for other factors. These data lend credence to the notion that the practitioner’s framework for mobilizing collective action that I introduced in this study is an effective tool for designing intervention strategies that can resolve the biggest policy problems faced by modern society
Quantifying and Modeling Contexts for Sequential Recommendation
Get PDFRecommender systems help users discover products, content, or actions aligned with their preferences. Traditionally, these systems have used content-based and collaborative filtering approaches: content-based recommenders suggest items similar to those the user has liked, while collaborative filtering identifies patterns of similarity between users or items. Recently, sequential recommendation has introduced time and order into recommendations, recognizing that user interests can shift dynamically. Unlike traditional methods that treat preferences as static, sequential recommendations provide contextually relevant suggestions by incorporating sequential context, making recommendations personalized, timely and contextually relevant. Though recognized as central to sequential recommendation, sequential context itself has rarely been explored in depth. This dissertation focuses on key challenges in quantifying and modeling sequential context in recommendation systems. In the first part, we address the dynamic, hierarchical, and cascading nature of sequential context, proposing a novel framework called Higher Order Latent Interactional Context (HOLIC) for modeling sequential context. This approach combines a multi-layer recurrent neural network, an attention mechanism, and a sequence clustering module to capture the underlying dynamical properties of sequential context. The second part examines the unique characteristics of training data for sequential recommendations, particularly when using transformer-based masked language models (MLMs) for item representation, a common pretraining approach. We find that applying MLMs naively to sequential recommendation data, without addressing differences between user interaction sequences and natural language, can degrade and distort item relationship signals. We propose a method to recover and amplify these signals by augmenting training data with synthetic permutations that preserve local and global sequential context. In the final section, we present pioneering work on multi-modal contexts for sequential recommendation, leveraging the rise of multi-media data—such as images, text, audio, and video in modern applications. We observe that multi-modal interactions add significant complexity, with item relationships potentially unique to each modality, redundant across them, or arising from intermodal synergies. We define and systematically categorize these challenges, and propose two models for uncovering dynamic sequential relationships within and across modalities. The first model treats the user interaction sequence as a heterogeneous temporal graph (HTG) and defines meta-relations on the HTG that encode modality types as node types and connections between modalities as edge types. The meta-relations are used to parameterize message passing and attention mechanisms to distill complex dynamic inter-modal and intra-modal signals into rich contextualized item embeddings. The second model is a self-attention based multi-modal fusion framework with a novel sequence compression technique to manage the transformer’s quadratic time complexity, allowing us to effectively model long multi-modal sequences with text and image interactions
Latinos in Massachusetts Selected Areas: Marlborough
Get PDFThis is a publication of the Mauricio Gastón Institute for Latino Community Development and Public Policy