A lack of distinct cell identities in single-cell measurements: revisiting Waddington’s landscape

Abstract

AbstractA prevailing interpretation of Waddington’s landscape is that distinct cell types with distinct physiologies are produced and stabilized by dynamical attractors in gene expression space. This notion is often applied in the analysis of single-cell omics data, where cells are clustered into groups that represent cell types, prior to downstream analyses like differential gene expression. Until the advent of single-cell measurement technologies, however, it has been impossible to characterize the heterogeneity of cells in the neighborhoods of these attractor states. In this work, we apply graph theory to characterize the distribution of cells in epigenetic space, using data from various tissues and organisms as well as various single-cell omics technologies. Rather than finding distinct clusters of cells that map cell types to specific regions of epigenetic space, we found that cells of very distinct types and lineages occupy the same region of space. Further, we found that the density distribution of cells is approximately power-law, with most cells existing in low-density regions, very far from other cells. This highly heterogeneous density distribution is unexpected, as it is not consistent with the distributions we would expect to see in the neighborhood of an attractor. We found these two observations are universal in single-cell data on epigenetic state of multicellular organisms, regardless of the tissue, organism, measurement technique employed, or the approach used to select the subset of genes on which the analysis was performed. The fact that currently-available single-cell data is inconsistent with the predictions of Waddington’s landscape poses a challenge both for the robust analysis of these data and for our overall understanding of epigenesis in development.