Fig. 4From: InterDILI: interpretable prediction of drug-induced liver injury through permutation feature importance and attention mechanisma AUROC and b AUPR performance of hold-out validation. Four data sets, NCTR, Greene, Xu, and Liew, were used as the training set and the DILIrank was only used as the test set. c AUROC and d AUPR performance of stratified tenfold cross validation. The shaded region indicates 95% confidence intervalBack to article page