Harnessing Biological Variability for Mechanistic Inference: A Practical Stochastic Framework

Abstract

Inter-individual heterogeneity is often treated as noise, yet its temporal evolution can reveal regulatory mechanisms hidden from mean-field behavior. We present a stochastic framework that exploits variability for mechanistic inference in cell population dynamics. Using adult neurogenesis as a case study, we develop a state-dependent stochastic model of transitions between quiescent and active states and derive a diffusion approximation for the dynamics of both mean and variance. Applied to repeated cross-sectional data from wild-type and interferon-receptor knockout mice, we show that distinct regulatory mechanisms can produce similar mean dynamics but different fluctuation patterns. Jointly fitting mean and variance identifies proliferation-rate regulation as the dominant contributor to variability, while activation and self-renewal primarily govern average and long-term dynamics. Wild-type mice exhibit regulation of all three processes, whereas knockout mice lose activation control. These results show that population-level variability provides mechanistic information beyond average dynamics and helps distinguish between competing mechanistic models.