Title: Exploring the space of genes in single cell transcriptomics datasets
Abstract: Single cell transcriptomics is a revolutionary technique in biology that allows for the measurement of gene expression levels across the genome for many individual cells simultaneously. Analysis of these vast datasets reveals variations in expression patterns between cells that were previously out of reach. On top of discrete clustering into cell types, continuous patterns of variation become visible, which are associated to differentiation pathways, cell cycle, response to treatment, adaptive heterogeneity or what just whatever the cells are doing at that moment. Current methods for assigning biological meaning to single cell experiments relies on predefining groups of cells and computing what genes are differentially expressed between them. The complexity found in modern single cell transcriptomics datasets calls for more intricate methods to biologically interpret both discrete clusters as well as continuous variations. We propose topologically-inspired data analysis methods that identify coherent gene expression patterns on multiple scales in the dataset. The multiscale methods consider discrete and continuous transcriptional patterns on equal footing based on the mathematics of spectral graph theory. As well as selecting important genes, the methodology allows one to visualise and explore the space of gene expression patterns in the dataset.