Figure 2.

Toy example to illustrate limitations of a two-step approach. The schematic demonstrates why clustering or biclustering followed by gene-enrichment analysis may miss patterns that can be found by our approach. The left panel shows two hypothetical experiments side by side using the same type of visualization as in Fig. 1. Genes with a particular functional annotation are highlighted; others are represented by dashed lines. It can be seen that the natural clusters in the data set do not show enrichment and would, hence not be considered significant by a two-step approach. In contrast, for our approach the histogram of neighbors (right panel) differs from what would be expected for a random subset.