Fig. 4. Pairwise enrichment.

The heatmaps show how the observed frequency of amino acid residues in two randomised positions ($f_{a,b}^{obs}$) divided by the expected frequency in an ideally balanced library ($f_a^{id}\cdot f_b^{id}$). A logarithmic scale is used to display both enrichment (red) and depletion (blue). Each panel represents one pair of randomised positions. The presence of large hydrophobic residues Leu/Ile at any one of positions 7a, 9 or 11 serves as an “anchor”, so that these residues appear in combination with non-preferred amino acids. n = 29,563 for each individual heatmap.