Figure 5.

Comparison of ROC curves for drug classification using DrugMetric, QED and ADMET-score across various dataset pairings. (A) DrugBank vs. ChEMBL, where DrugMetric achieved the highest AUC of 0.88, followed by QED with 0.79 and ADMET-score with 0.65. (B) ChEMBL versus WITHDRAW, with DrugMetric recording an AUC of 0.87, QED at 0.79 and ADMET-score at 0.70. (C) DrugBank vs. WITHDRAW, showing DrugMetric with an AUC of 0.58, QED at 0.52 and ADMET-score at 0.60. These results demonstrate the varying effectiveness of each scoring system in distinguishing drug-like molecules in different dataset contexts.