Learning geometric models for developmental dynamics

Abstract

AO_SCPLOWBSTRACTC_SCPLOWSuccessful development from a single cell to a complex, multicellular organism requires that cells differentiate in a coordinated and organized manner in response to a number of chemical morphogens. While the molecular underpinnings may be complex, the resulting phenomenon, in which a cell decides between one fate or another, is relatively simple. A body of work--rooted in dynamical systems theory--has formalized this notion of cellular decision making as flow in a Waddington-like landscape, in which cells evolve according to gradient-like dynamics within a potential that changes shape in response to a number of signals. We present a framework leveraging neural networks as universal function approximators to infer such a parameterized landscape from gene expression data. Inspired by the success of physics-informed machine learning in data-limited contexts, we enforce principled constraints motivated not by physical laws, but by this phenomenological understanding of differentiation. Our data-driven approach infers a governing landscape atop a manifold situated within expression space, thereby describing the dynamics of interest in a biologically meaningful context. The resulting system provides an intuitive, visualizable, and interpretable model of cellular differentiation dynamics.