Skip to main content

Table 7 Performance of the RF models on the two test sets when the cutoff is 20%

From: HobPre: accurate prediction of human oral bioavailability for small molecules

Data set Model SE SP ACC AUC MCC F1-score
Test set 1
(287)
Model 1 0.370 0.869 0.742 0.736 0.264 0.824
Model 2 0.479 0.864 0.767 0.767 0.360 0.847
Model 3 0.452 0.893 0.780 0.771 0.379 0.858
Model 4 0.452 0.856 0.746 0.759 0.308 0.832
Model 5 0.493 0.855 0.763 0.789 0.359 0.833
Consensus model 0.493 0.925 0.815 0.801 0.473 0.882
Test set 2
(133)
Model 1 0.8 0.891 0.887 0.973 0.384 0.938
Model 2 1 0.859 0.865 0.978 0.432 0.924
Model 3 0.8 0.883 0.880 0.939 0.371 0.933
Model 4 1 0.914 0.917 0.955 0.534 0.955
Model 5 1 0.898 0.902 0.973 0.499 0.946
Consensus model 1 0.906 0.910 0.981 0.516 0.951
  1. Bold numbers refer to the maximum value (optimum value) obtained from the corresponding evaluation index