A trade-off in controlling upstream and downstream noise in signaling networks

Signal transduction, underpinning the function of a variety of biological systems, is inevitably affected by fluctuations. It remains intriguing how the timescale of a signaling network relates to its capability of noise control, specifically, whether long timescale can average out fluctuation or accumulate fluctuation. Here, we consider two noise components of the signaling system: the upstream noise from the fluctuation of the input signal and the downstream noise from the stochastic fluctuations of the network. We discover a fundamental trade-off in controlling the upstream and downstream noise: a longer timescale of the signaling network can buffer upstream noise, while accumulate downstream noise. Moreover, we confirm that this trade-off relation exists in real biological signaling networks such as a fold-change detection circuit and the p53 activation signaling system. Author SummaryInformation transmission is vital in biological systems, such as decoding the information regarding nutrient levels during chemotaxis or morphogen concentrations in tissue development. While fluctuations arising from the stochastic nature of biological processes inevitably affect information transmission, noise control mechanisms have been studied for decades. However, it remains controversial what the role of the slow dynamics (long timescale) is in noise control. On the one hand, it has been reported to attenuate noise by averaging out fluctuations. On the other hand, it is also proposed to amplify noise by accumulating fluctuations. Here we dissect the noise in signaling systems into two components: upstream noise originating from signal fluctuation, and downstream noise from the network stochasticity. Our analysis reveals that upstream noise negatively correlates with timescale, while downstream noise exhibits a positive correlation, indicating a fundamental trade-off in controlling the two noise components. Moreover, we provide an intuitive illustration to understand this phenomenon using the concept of landscape representation. Mathematically, we analytically derive a trade-off relation that agrees well with simulations. Our results uncover a new property of noise in signaling processes, deepening our understanding of noise control and proposing a new perspective in designing signaling network.