Fig 4. Analysis of the parameter fitting.

(A): Top row: sample histograms of the parameters q_R, q_C and w_C. Yellow represents the entire sample population (uniform distribution), orange the samples for which the distances Φ^C and Φ^H are below the threshold τ = 0.05, and blue the samples fitting the entire truth table. Lower row: correlation between pairs of parameters in the 3 cases (same color code). For all 3 parameters, the orange histograms are no longer uniform, but restricted to smaller ranges. Such ranges concentrate even further for the feasible samples (blue), in particular the interval for q_C becomes quite tight. The correlation plots indicate that the boundaries between parameters subsets are well-defined and sharp. In particular, both q_R and w_CR have to be big enough in order to fulfill the entire truth table (i.e., blue points are in the top right corner). Notice how instead the binding affinity q_C cannot be big. (B): Sample histograms of the parameters q_A and q_B, and their correlation. Notice the sharp difference in the two histograms: q_A ≪ q_B for feasible samples. (C): Sample histograms of the interaction coefficients $w_{A_1{A}_2}$ and $w_{B_1{B}_2}$. The two orange histograms have a neat difference, which is however only partially reflected in the feasible samples (blue). (D): Sample histograms of the interaction coefficients $w_{A_1{B}_2}$ and $w_{B_1{A}_2}$. No clear trend appears.