Fig. 5: Effect of density ρq of transcription factor (TF)-otherTF interactions on stability.
a In the real network analyzed, there are us = 134 self-regulations (of which 42 of them are activating), ut = 373 TF-other TF interactions (of which 201 of them are activating), and un = 5148 TF-nonTF interactions (of which 2944 of them are activating). The total number of interactions is given by u. b A randomly constructed network is almost always stable when ρq is sufficiently low. Above a threshold value, the probability of being stable (P(stable)) decreases with ρq. This is true with (red and green circles) or without (blue circles) self-interactions, and regardless of whether it is the total number of interactions u (red circles) or the number of TF-nonTF interactions un (green and blue circles) that is kept constant. Each data point is an average over 15 sets of 10 regulatory networks, with error bars indicating the interquartile range. [Parameters: N = 2274, q = 211, h = 2, \({{{\Omega }}}_{\max }=1000\).].
