Figure 3.

Benchmarking the ability of the gene-gene co-expression matrices to predict relevant genes for terms within libraries for (A) Caenorhabditis elegans, (B) Danio rerio, (C) Drosophila melanogaster and (D) Saccharomyces cerevisiae. To compare these results to a baseline, AUCs were calculated after randomly shuffling terms. Benchmarking was also performed after removing half of the most redundant gene sets from each library in order to demonstrate robustness of these predictions to this factor.