Fig. 5
From: Accurate and efficient target prediction using a potency-sensitive influence-relevance voter
Simulated target-prediction experiment when training with random negatives: AUC scores as dataset size grows. Average AUC (y axis) plotted as a function of the minimum number of training molecules (x axis). Each method’s ability to separate known actives from a background set of 9000 random ChEMBL molecules, assumed to be inactive, is measured. 1000 random negative molecules are added to the original training sets. The extended training sets result in significant performance improvements