Figure 4.

Interaction scores for each protein. Each column shows the interaction scores using the base-optimized weights for one sequence's strong interactions, shown in red, and non-interactions, shown in blue. The 58 sequences are grouped by similarity, and ordered as in Table 1 of Additional data file 1. The blue horizontal lines mark off high-confidence predictions of interactions (more than 32.8) and non-interactions (less than 27.8). Note that all heterodimer interactions appear twice on the graph.