Fig. 2

Overall prediction performance by the two models analyzed in this work. a-d Logic behaviour of the circuits in terms of RFP output level in vivo (yellow bars) and in silico (cyan bars) in absence of HSL and at the maximum HSL concentration tested. Results are shown for training set (a-b) and test set circuits (c-d), considering NBM (a, c) and BM (b, d). Red squares surrounding the sub-panels indicate a circuit configuration with unexpected in vivo behaviour. Red edges in the in silico-predicted output bars indicate that the model is not able to predict the observed logic behaviour of the circuit. Yellow bars represent the average output value and error bars represent the 95% confidence intervals of the mean. Cyan bars represent the median predicted value and error bars represent the 95% confidence intervals calculated via Monte Carlo simulations. e-f Measured output of the circuits at all the HSL concentrations tested in this work plotted against the values predicted by the NBM (e) and BM (f). Red and blue points represent RFP and GFP output, and are expressed as AU_R cell⁻¹ min⁻¹ and AU_G cell⁻¹ min⁻¹, respectively. Asterisks correspond to the data of the two circuits showing unexpected in vivo behaviour, while circles correspond to the data of all the other test set circuits. The solid line is the bisector line. Each point represents the average value of the in vivo measured condition, versus the median value of the corresponding model prediction