Figure 2

Interface mutual information (\(\hat{I}_{AB}\)) optimization trajectories. (A) Scheme showing \(\hat{I}_{AB}\) optimization process starting from a scrambled multi-sequence alignment (MSA) concatenation (in gray) and reaching an optimized concatenation (in blue). Only physically coupled MSA position pairs (shown in purple) are taken into account. (B) Optimization trajectories for 26 protein systems. For each system, there are six trajectories with different starting points. The \(\hat{I}_{AB}\) normalized by the native interface mutual information (relative \(\hat{I}_{AB}\)) is plotted against the number of generations of the genetic algorithm (gray lines). The average trajectory over all complexes is shown in black. (C) First-order derivative of the optimization trajectories shown in (B). The derivatives of individual trajectories are shown in gray, while the average derivative over all trajectories is shown in black. This figure was generated with Inkscape (https://inkscape.org/) and matplotlib v3.1.2 (https://matplotlib.org/).