Fig. 2: Holimaps for autoregulatory gene networks in steady-state conditions. | Nature Communications

Fig. 2: Holimaps for autoregulatory gene networks in steady-state conditions.

From: Holimap: an accurate and efficient method for solving stochastic gene network dynamics

Fig. 2

a Stochastic model of an autoregulatory feedback loop, which includes bursty protein synthesis, protein decay, cooperative binding of protein to the gene, and unbinding of protein. b The LMA maps the nonlinear network to a linear one with effective parameter \({\hat{\sigma }}_{b}\). The high-order reactions G + hPG* in the former are replaced by the first-order reactions GG* in the latter. c The 2-HM maps the nonlinear network to a linear one with effective parameters \({\tilde{\sigma }}_{u}\) and \({\tilde{\sigma }}_{b}\). d The 4-HM maps the nonlinear network to a linear one with effective parameters \({\bar{\sigma }}_{u},{\bar{\sigma }}_{b},{\bar{\rho }}_{u}\), and \({\bar{\rho }}_{b}\). e Heat plot of the HD for the LMA as a function of the protein burst frequencies ρu and ρb. Here the HD for the LMA represents the Hellinger distance between the real steady-state protein distribution computed using FSP applied to the nonlinear system and the approximate protein distribution computed using the LMA. f Heat plots of the HDs for the LMA and Holimaps as functions of the unbinding rate σu and binding rate σb (normalized by the decay rate d) when ρbρu. The red curves enclose the true bimodal region, i.e., the parameter region in which the protein number has a bimodal distribution, as predicted by FSP; the orange curves enclose the bimodal region predicted by the approximation method. The vertical white dashed line demarcates the region of σu ≥ d, where the linear network given by the LMA can never exhibit bimodality, from the region of σu < d where it can. g Comparison of the steady-state protein distributions computed using FSP, LMA, 2-HM, and 4-HM in different regimes of gene state switching. h The maximum HD as a function of the cooperativity h for the LMA and Holimaps. Here the maximum HD is computed when σu and σb vary over large ranges, while other parameters remain fixed. See Supplementary Note 1 for the technical details of this figure. Source data are provided as a Source Data file.

Back to article page