Regulatory stochasticity drives opposing phenotypic outcomes in cell-fate decision networks

Abstract

Models of gene regulatory networks (GRNs) underlying cell-fate decision systems treat regulatory parameters as fixed quantities, although transcription-factor efficacy varies with molecular context. We asked how temporal fluctuations in interaction strength, rather than molecular abundance, alter phenotype occupancy in multistable regulatory networks. We applied three stochastic update rules to fold-change parameters—independent fluctuations anchored to their initial values, a bounded additive random walk, and a bounded multiplicative random walk—across RACIPE ensembles of mutually inhibitory circuits, and tested analogous perturbations in Boolean models of epithelial–mesenchymal plasticity and gonadal-fate determination. Anchored fluctuations largely preserved deterministic occupancies of network states. Additive fluctuations increased occupancy of all-high co-expression states, particularly in multistable regimes and when high expression was readily accessible; this effect extended to hybrid team-expression states in both biological networks. Multiplicative fluctuations biased inhibitory fold changes toward stronger repression and favored single-high states in a topology-dependent manner. Fixed-parameter controls sampled from the same noise-induced distributions did not fully reproduce the additive effect or the multiplicative effect in several circuits. Our analysis traced the additional additive bias to an asymmetry between unconditional derepression and conditional repression. Thus, temporally evolving regulatory parameters can redistribute phenotype occupancy in ways not predicted from static parameter ensembles, with outcomes jointly determined by fluctuation dynamics and network topology.