Fig. 2

The effect of the number of initial conditions on the convergence of the RACIPE results. a For each coupled toggle-switch I (CTS-I) circuit (curves in different colors), the convergence is evaluated by the dissimilarity of states using different numbers of initial conditions (nIC in x-axis) and different numbers of RACIPE models (nRM in different panels). b The minimum nIC to get the converged distribution of the number of stables states when nRM equals 10,000. Different points represent the CTS-I circuits of different sizes. The minimum nIC is selected if the decrease of the Bhattacharyya distance is smaller than the threshold (0.0005, see Additional file 1: Figure S3) when nIC increases. c For each CTS-I circuit, the convergence is alternatively evaluated by the dissimilarity of expressions of each gene. Only the Ai genes for each circuit are plotted (one line per gene) and colored differently for different nRMs. The dissimilarity is less sensitive to nIC, but is dramatically reduced with the increase of nRM. d The minimum nIC to get the converged distribution of expressions. The minimum nIC is selected if the decrease of the Bhattacharyya distance is smaller than the threshold (0.0005, see Additional file 1: Figure S6) when nIC increases. nRM needs to be larger than 5000, otherwise the distribution is not converged even with nIC = 2000