Statistical and Computational Methods for High-Dimensional Genomics Data
dc.contributor.author | Ma, Ying | |
dc.date.accessioned | 2023-09-22T15:18:36Z | |
dc.date.available | 2023-09-22T15:18:36Z | |
dc.date.issued | 2023 | |
dc.date.submitted | 2023 | |
dc.identifier.uri | https://hdl.handle.net/2027.42/177705 | |
dc.description.abstract | Advancements in transcriptomic technologies have enabled the measurement of gene expression at single cell resolution and provided spatial localization information on tissues. The increasing accessibility of these single-cell RNA sequencing (scRNA-seq) or spatially resolved transcriptomic (SRT) datasets provides a comprehensive cell atlas. It enables the thorough characterization of transcriptomic landscapes of tissues for a mechanistic understanding of many biological processes. In the meantime, improvements in transcriptomic technologies have increased both the volume and complexity of data, introducing new computational and statistical challenges for data analysis, including differential expression analysis, gene set enrichment analysis, cell type deconvolution analysis, and spatial domain clustering. In this dissertation, I propose three statistical and computational methods to address these challenges for capturing and dissecting cellular and tissue heterogeneity with high statistical power and accuracy, while providing new insight into biological systems. In Chapter 2, I develop a method, iDEA, that performs joint DE and GSE analysis in scRNA-seq studies. By integrating DE and GSE analyses, iDEA can improve the power and consistency of DE analysis, produce effective control of type I errors, thus yielding high statistical power and accuracy of GSE analysis. Importantly, iDEA uses only DE summary statistics as input, enabling effective data modeling through complementing and pairing with various existing DE methods. I illustrate the benefits of iDEA with extensive simulations, and three scRNA-seq data sets, where iDEA achieves up to five-fold power gain over existing GSE methods and up to 64% power gain over existing DE methods. In Chapter 3, I develop a method CARD to perform spatially informed cell type deconvolution for SRT data. CARD builds upon a non-negative matrix factorization (NMF) model that leverages the cell-type-specific gene expression from scRNA-seq data. A unique feature of CARD is its ability to accommodate the spatial correlation structure in cell-type composition across tissue locations by a conditional autoregressive (CAR) modeling assumption. This enables accurate and robust deconvolution of SRT data across technologies and in the presence of mismatched scRNA-seq references. Furthermore, modeling spatial correlation allows CARD to impute cell-type compositions and gene expression levels on new locations of the tissue, facilitating the reconstruction of high-resolution map. Importantly, CARD is computationally scalable and efficient to datasets with tens of thousands of genes measured on tens of thousands of samples. With extensive simulations and comprehensive applications to four real datasets, CARD outperforms other methods, provide novel biological insight underlines the tissue heterogeneity. In Chapter 4, I develop a method that simultaneously characterize the transcriptomic landscapes on multiple tissues. While SRT datasets can be generated from multiple tissue sections with high resolution, existing methods primarily focus on a single tissue section and fail to utilize information from scRNA-seq datasets for spatial domain detection. Additionally, many published methods lack computational scalability for high-resolution large-scale SRT datasets being collected today. To fill these gaps, I developed IRIS, which leverages cell type specific gene expression information from scRNA-seq to detect spatial domains on multiple tissue sections. By iteratively updating spatial domain labels while considering within-slice and between-slice compositional similarities, IRIS ensures optimal clustering performance. Through in-depth analysis of six spatial transcriptomics datasets, IRIS demonstrates significant advantages, achieving up to 1083% clustering accuracy improvement over existing methods. This enables the identification of transcriptomic landscapes in complex tissues, including the human prefrontal cortex, spermatogenesis, olfactory bulb, and human breast cancer. | |
dc.language.iso | en_US | |
dc.subject | Statistical Methods | |
dc.subject | Single-cell RNA-seq | |
dc.subject | Spatial Transcriptomics | |
dc.title | Statistical and Computational Methods for High-Dimensional Genomics Data | |
dc.type | Thesis | |
dc.description.thesisdegreename | PhD | en_US |
dc.description.thesisdegreediscipline | Biostatistics | |
dc.description.thesisdegreegrantor | University of Michigan, Horace H. Rackham School of Graduate Studies | |
dc.contributor.committeemember | Zhou, Xiang | |
dc.contributor.committeemember | Welch, Joshua | |
dc.contributor.committeemember | Fritsche, Lars | |
dc.contributor.committeemember | Morrison, Jean | |
dc.subject.hlbsecondlevel | Science (General) | |
dc.subject.hlbtoplevel | Science | |
dc.description.bitstreamurl | http://deepblue.lib.umich.edu/bitstream/2027.42/177705/1/yingma_1.pdf | |
dc.identifier.doi | https://dx.doi.org/10.7302/8162 | |
dc.identifier.orcid | 0000-0003-3791-7018 | |
dc.identifier.name-orcid | Ma, Ying; 0000-0003-3791-7018 | en_US |
dc.working.doi | 10.7302/8162 | en |
dc.owningcollname | Dissertations and Theses (Ph.D. and Master's) |
Files in this item
Remediation of Harmful Language
The University of Michigan Library aims to describe library materials in a way that respects the people and communities who create, use, and are represented in our collections. Report harmful or offensive language in catalog records, finding aids, or elsewhere in our collections anonymously through our metadata feedback form. More information at Remediation of Harmful Language.
Accessibility
If you are unable to use this file in its current format, please select the Contact Us link and we can modify it to make it more accessible to you.