Maximized redundant and synergistic information transfers predict the rise in the output gene expression noise in a generic class of coherent type-1 feed-forward loop networks

We apply the partial information decomposition principle to a generic coherent type-1 feed-forward loop (C1-FFL) motif with tunable direct and indirect transcriptional regulations of the output gene product and quantify the redundant, synergistic, and unique information transfers from the regulators to their target output species. Our results which are obtained within the small-noise regime of a Gaussian framework reveal that the redundant and synergistic information transfers are antagonistically related to the output noise. Most importantly, these two information flavors are maximized prior to the minimization and subsequent growth of the output noise. Therefore, we hypothesize that the dynamic information redundancy and synergy maxima may possibly be utilized as efficient statistical predictors to forecast the increasing trend of the fluctuations associated with the output gene expression dynamics in the C1-FFL class of network motifs. Our core analytical finding is supported by exact stochastic simulation data and furthermore validated for a diversified repertoire of biologically plausible parameters. Since, the output gene product serves essential physiological purposes in the cell, a predictive estimate of its noise level is supposed to be of considerable biophysical utility.