Speaker
Description
Cells respond to their environment through protein networks often dysregulated in cancer, making predictive modelling crucial. Because experiments capture only limited observables, coarse-graining is needed to uncover low-dimensional descriptions. Yet classical approaches rely on idealised assumptions, leaving it unclear when partial experimental observations support reduced system dynamics. Here we show that symbolic regression (SR) provides a principled framework to probe the existence of coarse-grained dynamics and, when they exist, infers mechanistically interpretable models. In synthetic enzyme systems, SR recovers Michaelis–Menten kinetics for two- and three-step mechanisms, but not for a four-step dimerisation model. As data quality is degraded, SR simplifies toward effective kinetic laws while preserving correct theoretical limits. Applied to published time-resolved ERK phosphorylation data, SR identifies compact non-linear ERK rate laws in selected cancer-relevant overexpression contexts, yielding interpretable kinetic effects when models are predictive, and otherwise failing alongside our Neural ODE baseline, indicating missing observations. Together, these findings establish a data-driven approach to identifying coarse-grained models of signalling. More broadly, they establish symbolic regression as a framework for guiding experiment design by generating hypotheses where reduced laws emerge and motivating new measurements where they do not.
