Gene-free landscape models for development
Cell fate decisions in developing tissues involve gene regulatory networks that often comprise multiple genes, many molecular components and elaborate signalling dynamics. Despite the complexity, the outcomes of cellular decisions are relatively simple: cells transition between a limited set of discrete cell fates, each defined by a distinct gene expression profile. A popular and intuitive metaphor for the process of developmental decision making is the Waddington landscape, in which the differentiation trajectory of a cell is conceived as a ball rolling down a landscape of branching valleys. This can be mathematically formalised using dynamical systems theory where the landscape is defined by a potential function and the different cell states correspond to attractors in the landscape. In this setting, variation in the parameters of the dynamical system, caused by changes in the signals the cell receives, alter the landscape and give rise to bifurcations that destroy or create attractors. We took advantage of the differentiation of neural and mesodermal cells from pluripotent progenitors using mouse embryonic stem cells (ESCs) exposed to different combinations and durations of signalling factors to build a dynamical landscape that reproduces the training experimental data and allows for predictions of untested experimental conditions.