Bayesian Information Processing in Stochastic Intracellular Systems

Abstract

Biological systems can operate robustly even with substantial stochasticity in their components. One possible but not yet proven mechanism to implement robust operation with noisy components is that relevant information for robust control is embedded in apparently stochastic signals. In this work, by employing Bayesian theory, we theoretically show that intracellular reactions with specific structures can implement statistically optimal dynamics to decode (extract) the relevant information embedded (encoded) within the apparently noisy signal. We also demonstrate that the decoding dynamics is related to a noise-induced transition, implying that optimal dynamics to suppress noise behaves as if exploiting noise for signal amplification.