Computational analyses of HTseq data from developing B cells

B lymphocytes are critical components of the adaptive immune system, equipped with the ability to generate a widely diverse repertoire of antigen receptors through V(D)J recombination, to undergo affinity maturation in germinal centers, and to differentiate into antibody-secreting plasma cells and long-lived memory B cell populations. Their complex functions are influenced by both systemic physiological contexts and the microenvironmental cues of lymphoid organs. Advances in high-throughput sequencing technologies, particularly bulk and single-cell RNA sequencing, have enabled a deeper understanding of these processes by unraveling the details of molecular interactions of immune cells at an unprecedented scale and resolution, at the expense of significant computational challenges, novel technical artefacts, interpretational complexities, and reproducibility concerns. This dissertation brings together experimental immunology and computational approaches to address open questions in B cell biology and single-cell transcriptomics. The first two chapters aim to introduce essential concepts of B cell development and function (chapter 1), and the principles of high-throughput transcriptomics, in both bulk and single-cell contexts, and their analysis workflows (chapter 2). Chapter 3 presents the key findings of this dissertation and in chapter 4 a discussion on the three contributed articles concludes the present work. The first publication (section 3.1) illustrates how ischemic stroke and myocardial infarction induce profound metabolic dysfunction and stress responses in intestinal B cells, by triggering the release of neutrophil extracellular traps, which cause rapid loss of mucosal B cells, shrinkage of Peyer’s patches, and depletion of IgA-producing plasma cells. A bulk RNA-seq analysis of intestinal B cells after stroke revealed an enrichment in pro-apoptotic and stress-regulatory genes and an upregulation of DNA repair pathways, while metabolic and cell function genes were reduced and mitochondrial, ribosomal, and metabolic pathways were downregulated. In the second publication (section 3.2), scBubbletree is introduced; a new computational method for visualizing single-cell RNA-seq data that can easily scale up to even millions of cells. scBubbletree groups transcriptionally similar cells into robust clusters and displays them as "bubbles" arranged on a dendrogram, with bubble sizes reflecting cell counts and bubble positions depicting the structure of the dataset and the relationships between clusters. This approach avoids overplotting, preserves both local and global transcriptional relationships better than two-dimensional embeddings, and provides quantitative summaries of cluster attributes and metadata, such as cell type abundances or marker gene expression averages. scBubbletree is implemented to integrate with standard single-cell analysis workflows, ease comparison across conditions and modalities, and streamline with downstream functional analyses, while providing a more interpretable visualization of the data. The third publication (section 3.3), addressed a critical technical artefact in droplet-based single-cell transcriptomics, multiplets, which are formed when two or more cells are encapsulated in the same droplet. This study quantified the prevalence of multiplets across various datasets using cell hashing as a lower-bound ground truth and a refined Poisson-based model as an estimate for the true number of multiplets. Multiplet contamination was systematically underestimated and the non-uniform distribution of multiplets across the dataset was shown to be a significant confounder in downstream analyses, by distorting clustering, generating artificial clusters, and bias differential expression analysis, highlighting the need for rigorous analytical workflows and more robust multiplet-detection strategies.