(A) We select a target gene (green) and TFs (blue), which are known from the literature to interact with the target gene. We consider only target genes and corresponding TFs, which show significant variation in our dataset. (B) The fitting task is to find a regulation model using the TF mRNA expression level (upper plot), so that the time series of the target synthesis rate (dots in lower plot) is best described by the model synthesis rate (solid curve in lower plot). (C) We describe the concentration of active TF protein by a simple model with translation rate and linear degradation rate (see box quantitative model). For a correctly chosen the target gene synthesis rate plotted against the TF protein concentration collapses to a curve (bottom plot). (D) Together with a regulation function (target synthesis rate as a function of TF protein ) is estimated. For a single TF the regulation function has four parameters () and two possible directions of regulation: activator and repressor (see box mathematical model). To find the right regulatory direction we fit both and select the one, which yields the better score. (E) Two TFs can interact in multiple ways to generate a two dimensional combinatorial regulation function (Buchler et al., 2003). For two TFs we estimate the protein models and six parameters for the regulation function. There are 10 non-trivial regulatory analog 'logic' operations to test, of which 3 examples are depicted (see box quantitative model for the corresponding equations).