Master of Science Capstone program students partner with regional health organizations to complete their culminating capstone project.
Working in a small group with the project’s sponsor, students develop tailored solutions for multiple questions and needs.
Here are examples of recent projects:
2023 MS Capstone Projects
Benchmarking Cell Typing Methods in Spatial Transcriptomics
- Sponsor: NanoString Technologies Inc
- Team: Howard Baek, Eliza Chai, Alexis Harris, Ingrid Luo, and Makayla Tang
In 2020, Nature Methods crowned spatially resolved transcriptomics Method of the Year (Nature, 2021). Spatially resolved transcriptomics has changed the landscape of RNA sequencing and cell type analyses. With spatially resolved transcriptomics, researchers and scientists can retrieve cell transcriptomic data and locate the positional context of those cells in tissue (Nature, 2021). A critical component in analyzing spatial transcriptomic data is cell typing, which is the process of determining for every cell in the dataset, “what kind of cell is this?” NanoString Technologies, Inc., has developed a cell typing algorithm, Insitutype. Insitutype is designed for statistical and computational efficiency for spatial transcriptomic data. For this capstone, we ran competing cell typing methods against Insitutype on two different datasets and scored the accuracy of these models. Additionally, we performed a benchmarking analysis to determine how the performance of Insitutype varies based on tuning parameters.
A novel approach to analyzing the caliber of University of Washington Medicine clinical ethics consults
- Sponsor: UW Medicine Clinical Ethics Consult Service
- Team: Alison King and Katie McFarlane
Partnering alongside the UW Department of Bioethics and Humanities, we created a novel method for ranking clinical ethics consultants in terms of consultation completion and quality, as determined by bioethicist expertise and a pre-existing quality standard created by the United States Department of Veterans Affairs National Center for Ethics Care. These findings will be used to improve the consultations at the UW Medical Centers researched within this study and will hopefully be applied more broadly to other hospitals in order to improve consultations and better serve the clients utilizing these services. Our analysis serves as a first step towards further quantitative research and objective measurement in the field of clinical ethics.
Prediction of member engagement
- Sponsor: Community Health Plan of Washington
- Team: Thuy Nguyen, Jiayu Sun, Lester Tsai, and Lishan Wang
Community Health Plan of Washington (CHPW) is a non-profit health plan organization that provides care management services to their members. Identifying the members that will opt into CHPW’s care management programs is difficult, which leads to a waste of limited outreach resources. In this study, we build diverse machine learning models to predict member engagement in CHPW’s care management programs and evaluate their prediction performance. We obtain a prediction accuracy of about 0.62 for all models using classification threshold 0.4 which balances the tradeoff between precision (about 0.5) and recall (about 0.55). We prove that the use of our predictive models will not introduce racial bias. Moreover, we extract features that are important to the prediction of member engagement from our models, which can increase our understanding towards member engagement.
Transcriptomic Signatures Associated with Right Ventricular Failure and Outcomes in Pulmonary Arterial Hypertension
- Sponsor: UW Pulmonary, Critical Care, and Sleep Medicine
- Team: Hongyang Pi, Lan Shui and Runjia Zou
Partnered with UW Medical Center, Division of Pulmonary, Sleep and Critical Care Medicine to explore transcriptomic profiles associated with phenotypes of pulmonary arterial hypertension (PAH) in order to better understand the molecular mechanisms in PAH disease severity and right heart failure. Specifically, we analyzed the circulating RNA sequences in the plasma of PAH patients and discovered pathways and transcripts in mortality and right ventricular dilation in multivariate regression models. We demonstrated the utility of plasma transcriptomics in studying PAH and identified potential pathways for diagnostic and therapeutic targets.
Comparative analysis of the treatment of asymptomatic and symptomatic complex aortic aneurysms
- Sponsor: UW Vascular Surgery
- Team: Jennifer Ci, Thu Vu and Hanyi Wang
Partnered with Dr. Sara L. Zettervall at the UW Medicine Department of Surgery, Division of Vascular Surgery to identify the differences in outcomes for asymptomatic and symptomatic complex aortic aneurysm patients treated with fenestrated-branched endovascular aortic repair (F-BEVAR). Specifically, the analysis explored the relationship between presentation of symptoms (asymptomatic or symptomatic) and procedure outcomes such as survival, reintervention, ICU/hospital stay, and other post-operative factors. This information will help surgeons and physicians decide the best course of treatment and provide prognostics based on specific characteristics for their patients.
Comparison of self-reported physical activity and actigraphy-measured physical activity in older adults living with HIV
- Sponsor: UW Nursing
- Team: Walter Chen, Ziyi Chen, Han Liu and Yingying Wei
Description: We have collaborated with the Webel Research Lab at the UW School of Nursing to conduct a study that examines the self-reported physical activity levels measured by the LRCQ questionnaire and actigraphy-measured physical activity in individuals living with HIV (PLWH). Our goal was to determine the concordance between these two measures. We collected one-year physical activity data from 103 patients in the PROSPER-HIV study simultaneously and analyzed them using chance-corrected agreement methods. Our findings showed poor agreement between LRCQ responses and actigraphy measures, suggesting that caution is necessary when using these measures interchangeably to interpret studies measuring physical activity levels. Therefore, we recommend creating more precise and dependable questionnaires to measure physical activity levels. Additionally, careful consideration should be given when selecting measurement tools to guarantee accurate and reliable results.