17 November 2023
Eric Deeds
Department of Ingegrative Biology and Physiology
UC Los Angeles
For over 80 years, Waddington’s landscape has provided a paradigm for cell differentiation during development. In its modern interpretation, the landscape posits that different cell types correspond to different stable attractors in gene expression space. This notion is often applied in single-cell omics data, particularly single-cell (sc) RNA-seq data, where groups of cells are clustered into distinct cell-type groups and then subjected to downstream analysis. In this work, we applied graph theory to directly characterize the distribution of cells in epigenetic space, using scRNA-seq data from various tissues, organisms and platforms, as well as other single-cell omics technologies. We found that, rather than corresponding to distinct attractors on the epigenetic landscape, different cell types actually occupy exactly the same regions of epigenetic space. We also found highly heterogeneous density distributions in this space, with local densities varying across several orders of magnitude. These findings are inconsistent with the expected densities near an attractor. We showed that this lack of attractor structure could not be explained by technical noise, scale variance among genes, nor the subset of genes that were used; nor could it be rescued by any set of transformations that compose the "standard pipeline" in scRNA-seq analysis. These findings pose a challenge for the robust analysis of single-cell data and raise the possibility for alternative explanations of canalization during development.