Fig. 3

Simulation of gene-set clustering. a Three models for gene-set clusters. Dots represent genes and circles represent gene-sets. First model (Case 1) assumes that two subgroups (orange and sky-blue) are merged using MM distance but contain dense PPIs within each subgroup. Second model (Case 2) represents two subgroups having insufficient overlap to be merged using MM score, but containing dense PPIs across the subgroups. In the first and second models, only three gene-sets (instead of five) were represented in each subgroup to simplify the figures. In third model (Case 3), a large gene-set contains multiple small gene-sets having overlaps. Black and gray dots represent genes included in gene-set and background, respectively. Lines represent PPIs between genes. Clustering results for (b) Case 1 and (c) Case 2 are shown. Ratios in the y axes indicate the successful detection ratios. Results for Case 3 was not shown because all trials showed the same result for each distance measure (See the text). d Rank distributions of gene-set sizes of unclustered gene-sets in the imperfectly merged instances in Case 2. e A simulation for KAPPA where set A (n = 10) is contained in another set B (n = 10–200). KAPPA is strongly affected by the set-size