Item Details

title

COMBINING FLOW CYTOMETRY AND MACHINE LEARNING TO PREDICT OUTCOMES FOR ALLOGENEIC HEMATOPOIETIC STEM CELL TRANSPLANT PATIENTS

author

Short, Samantha

abstract

This study aims to build computational models that predict outcomes for allo-HSCT patients. We obtained blood samples from patients at seven different timepoints post-transplant and analyzed the samples via flow cytometry. This flow cytometry data consisted of both surface markers and cytokines to identify various immune cell populations and T cell function. Immune profiles were curated for each patient sample based on the percentage of each population in the sample. Our analysis showed a correlation between CD4+ memory T cells and acute GVHD (aGVHD); cytolytic T cells and chronic GVHD (cGVHD); and activated CD4+ T cells and relapse. We aimed to confirm the importance of these immune populations through their predictive ability. Previous studies were not able to use machine learning models for prediction because of their small sample sizes. The size of our cohort allowed us to divide our data into training and testing sets, and with repeated cross-validation, we could use our data to train our models multiple times. Each model uses either 5 or 15 predictive variables. Currently, our models are underperforming, but we aim to increase accuracy through different sampling techniques and additional samples.

subject

flow cytometry

graft versus host disease

hematopoietic stem cell transplant

immune profiles

machine learning

contributor

Grayson, Jason M (committee chair)

Khuri, Natalia (committee member)

date

2022-01-15T09:35:32Z (accessioned)

2024-01-14T09:30:07Z (available)

2022 (issued)

degree

Biomedical Science – MS (discipline)

embargo

2024-01-14 (terms)

identifier

http://hdl.handle.net/10339/99392 (uri)

language

en (iso)

publisher

Wake Forest University

type

Thesis