Fig. 4

GLADIATOR outperforms previous methods. Enrichment scores (–log(hypergeometric p value)) for disease-gene associations extracted from DisGeNET vs. all predicted disease-gene associations (a); percentage of enriched modules vs. corresponding disease gold-standard associations extracted from DisGeNET (b); percentage of enriched modules vs. known pathways extracted from MSigDB (c). GLADIATOR predictions were compared to the modules obtained from the DIAMOnD algorithm, using both the full list of disease-gene associations (DIAMOnD-Full seed) and the largest connected component obtained from this list, which served as the seed for our algorithm (DIAMOnD-LCC), and to the largest modules obtained from the HotNet2 algorithm (HotNet2-All) and a subset of these modules reported as significant by the HotNet2 algorithm (HotNet2-Significant)