PhD Defence

Abstract

Single-cell RNA sequencing (scRNA-seq) is a technology that allows for quantifying the abundance of RNA molecules in individual cells. By profiling thousands of cells in a single experiment, scientists are provided a bird’s-eye view of the transcriptome of the different cell types within a biological sample or tissue. Establishing how transcriptomes differ between biological groups, e.g., between cells from healthy and diseased subjects, is instrumental to both basic and translational research. Statistical methods that can infer on changes in the average RNA expression between biological groups, commonly referred to as differential gene expression (DGE) methods, thus play a crucial role in most scRNA-seq data analysis workflows. This PhD dissertation has aimed to add resolution to canonical DGE methods in several ways. Most DGE methods provide inference at the level of genes. However, due to posttranscriptional modifications like alternative splicing, multiple mature RNA molecule can be produced from a single genomic locus. Two projects in this dissertation have focused on shifting the biological resolution of DGE methods from genes to mature RNA molecules. Second, DGE methods typically only focus on identifying genes for which the average expression differs between biological conditions, neglecting differences in any other aspect of the expression distribution. Instead, we propose a workflow that allows for identifying changes both in average expression and the fraction of cells in which the gene is detected. Finally, most current DGE methods either ignore cell-cell correlation or sacrifice singlecell level resolution to implicitly address this correlation. Here, we evaluate a statistical framework, Generalized Estimation Equations, to explicitly model the hierarchical correlation structure of multi-patient scRNA-seq datasets.

Register by Tuesday, May 21st.