Cell Type Deconvolution Using Single-Cell Rna-Seq Data

Abstract

The advent of high-throughput sequencing technologies such as RNA-Seq has enabled researchers to study gene expression at a whole-transcriptome level. This has led to important breakthroughs in our understanding of the function of genes and their regulation in various biological contexts. However, RNA-Seq data has limitations, especially when it comes to analyzing samples at the single cell level. This is because gene counts are averaged within each sample, which ignores vital information regarding the underlying cellular heterogeneity. To address this issue, cell type deconvolution, also known as cell type decomposition, has been developed as a method to estimate the proportions of different cell types present in a sample of tissue. In this dissertation, two different cell type deconvolution methods are proposed - one for bulk RNA-Seq data and the other for spatial transcriptomics data. In the first project, a cell type deconvolution method, designed for bulk RNA-Seq data and based on an incomplete reference, has been developed and applied to analyzing human metastatic cancers. This method allows for the estimation of cell type proportions for each sample, even in cases where unknown cell types are present. By doing so, it helps to overcome limitations associated with bulk RNA-Seq data and enables more accurate analysis of gene expression patterns in different cell types. The second project focuses on a cell type deconvolution method aimed at spatial transcriptomics data. Spatial transcriptomics data provides detailed information on the expression levels of genes in different locations within a tissue sample. However, gene expression is also averaged within each location, which overlooks cell type-specific effects within and across locations. To overcome this limitation, SPADE is proposed to integrates spatial location and histology to identify spatial domain-specific cell type composition. This enables a more comprehensive understanding of the cellular organization and function within a tissue and can shed light on how gene expression patterns vary across different tissues and conditions. Overall, the development of these cell type deconvolution methods represents a significant step forward in our ability to analyze gene expression data at the single cell level. By enabling more accurate analysis of gene expression patterns in different cell types, these methods have the potential to improve our understanding of the underlying biological mechanisms that govern gene expression and their regulation.