Schriever, Hannah C
(2023)
Machine Learning Approaches to Analyzing scRNAseq Data with Applications to the Heart and Eye.
Doctoral Dissertation, University of Pittsburgh.
(Unpublished)
![[img]](https://d-scholarship.pitt.edu/style/images/fileicons/application_pdf.png) |
PDF (Updated Dissertation. Restricted until 2026 (2 years))
Updated Version
Restricted to University of Pittsburgh users only until 12 December 2025.
Download (189MB)
| Request a Copy
|
Abstract
Since its inception, the use of single-cell RNA sequencing has steadily grown because of its capacity to illuminate cellular heterogeneity. This technology produces large datasets with technical artifacts and potentially non-linear patterns. Therefore, computational tools designed to analyze single cell RNA sequencing (scRNA-seq) in addition to custom approaches developed specifically for complex datasets when existing methods fail are needed.
The overall theme of this dissertation is developing and applying machine learning methodologies to scRNA-seq data, with specific applications to heart and eye development. First I introduce vaeda, a computational tool for annotating doublets in scRNA-seq data, and show its competitive performance with existing methods. Vaeda serves as a generalizable tool, seamlessly integrating into Python workflows and capable of application to any sufficiently large scRNA-seq dataset. Then I discuss a hierarchical random forest model I developed to characterize novel heart organoid differentiation protocols. Heart organoids are generated through the three-dimensional differentiation of induced pluripotent stem cells, offering distinct advantages. Notably, organoids can be produced from adult human cells and modified to carry specific mutations, thereby enabling the exploration of early human cardiac development without the need for primary human fetal cells. My work showed their protocols not only generated organoids with atrial and ventricular identities, but also were capable of recapitulating congenital heart defects in vitro. Lastly I describe my contributions to investigating the role of hydroxymethylation, an epigenetic modification, in retina development. Zebrafish with a global loss of retinal hydroxymethylation have been recently engineered and profiled along with their sibling controls to generate time series scRNA-seq data. From this data, I inferred developmental trajectories to identify promising target genes for planned retina specific knockout experiments.
Altogether, the work in this dissertation pushes the field of scRNA-seq analysis forward by introducing a novel generalizable computational tool and developing tailored analysis approaches for developmental biology datasets.
Share
| Citation/Export: |
|
| Social Networking: |
|
Details
| Item Type: |
University of Pittsburgh ETD
|
| Status: |
Unpublished |
| Creators/Authors: |
|
| ETD Committee: |
|
| Date: |
12 December 2023 |
| Date Type: |
Publication |
| Defense Date: |
21 August 2023 |
| Approval Date: |
12 December 2023 |
| Submission Date: |
17 September 2023 |
| Access Restriction: |
2 year -- Restrict access to University of Pittsburgh for a period of 2 years. |
| Number of Pages: |
199 |
| Institution: |
University of Pittsburgh |
| Schools and Programs: |
School of Medicine > Computational Biology |
| Degree: |
PhD - Doctor of Philosophy |
| Thesis Type: |
Doctoral Dissertation |
| Refereed: |
Yes |
| Uncontrolled Keywords: |
single cell RNA sequencing
machine learning
retina development
heart development
organoids
doublet |
| Date Deposited: |
12 Dec 2023 20:52 |
| Last Modified: |
16 Sep 2024 18:56 |
| URI: |
http://d-scholarship.pitt.edu/id/eprint/45408 |
Metrics
Monthly Views for the past 3 years
Plum Analytics
Actions (login required)
 |
View Item |
![[feed]](/images/feed-icon-32x32.png){kind=icon}