Summarized results regarding the effect of the alignment method on alignment quality for (a) synthetic networks, (b) PPI networks, and (c) protein-GO networks. In panel (a), there are three considered alignment methods (WAVE, MAGNA++, and SANA). In panels (b) and (c), there are two considered alignment methods (WAVE and SANA). For each case (see below), we compare the alignment methods and rank the different methods from best (rank 1) to worst (rank 3 in panel (a), and rank 2 in panels (b) and (c)). Then, we compute the percentage of all cases in which the given method is ranked as the first, second, or third best among all considered methods. In panel (a), there are 2 networks (geometric, scale-free) × 5 noise levels (0, 10, 25, 50, 75) = 10 cases. In panel (b), there are 4 networks (APMS-Expr, APMS-Seq, Y2H-Expr, Y2H-Seq) × 5 noise levels (as above) = 20 cases. In panel (c), there are 2 networks (protein-GO-APMS, protein-GO-Y2H) × 5 noise levels (as above) = 10 cases. Note that we analyzed an additional noise level (100%), but we leave the corresponding results from this summary figure, because at this level all cases are expected to result in the same (random) alignments (Section Evaluation-Creating noise counterparts of a synthetic, PPI, or protein-GO network). Instead, we show the results for the noise level of 100% in the detailed figures (Figs 5, 6 and 7). Also, note that in this figure, we give each method the best case advantage. That is, we show results for the best of HetNC-HomEC and HetNC-HetEC, and also only for the maximum node color level (four colors in panels (a) and (b), and two colors in panel (c)). We do the latter because of all color levels, it is the maximum color level at which the given method performs the best, for each method. Nonetheless, the results remain qualitatively the same if we account for all considered colored levels.