Skip to main content

Table 2 Performance results for different feature sets

From: Using predicate and provenance information from a knowledge graph for drug efficacy screening

Feature set AUC Guney reference set AUC EMC reference set
Overlap and co-occurrence features 59.8% (0.9%)* 64.9% (0.6%)
Overlap and predicate features 77.6% (1.6%) 73.1% (0.9%)
Overlap and provenance features 75.1% (1.7%) 71.3% (1.0%)
Overlap, predicate and provenance features (all relationships) 78.1% (1.7%) 74.3% (1.0%)
Predicate and provenance features (indirect relationships only) 74.4% (1.9%) 70.6% (1.0%)
Guney’s proximity metric 65.6% (1.4%) 64.6% (0.6%)
  1. *Values indicate mean and standard deviation of the AUCs of 100 experiments