Skip to main content

Advertisement

Table 4 Validity and efficiency for active and inactive compounds at the 80% confidence level for the derived conformal predictors based on physicochemical descriptors

From: Maximizing gain in high-throughput screening using conformal prediction

AID Validity active Efficiency active Validity inactive Efficiency inactive
411 train 0.856 0.809 0.815 0.771
411 test 0.873 0.847 0.811 0.794
868 train 0.828 0.798 0.813 0.835
868 test 0.825 0.844 0.805 0.862
1030 train 0.823 0.654 0.819 0.636
1030 test 0.832 0.677 0.807 0.653
1460 train 0.864 0.864 0.816 0.88
1460 test 0.748 0.944 0.805 0.957
1721 train 0.868 0.918 0.842 0.899
1721 test 0.869 0.933 0.835 0.907
2314 train 0.813 0.81 0.807 0.808
2314 test 0.801 0.833 0.803 0.819
2326 train 1 0.395 0.856 0.144
2326 test 1 0.511 0.849 0.151
2451 train 0.884 0.746 0.836 0.66
2451 test 0.859 0.778 0.828 0.707
2551 train 0.819 0.916 0.809 0.906
2551 test 0.812 0.944 0.803 0.934
485290 train 1 0.51 0.86 0.15
485290 test 1 0.545 0.863 0.137
485314 train 0.846 0.762 0.824 0.726
485314 test 0.856 0.799 0.818 0.743
504444 train 0.833 0.749 0.813 0.755
504444 test 0.818 0.767 0.811 0.771
  1. Train denotes the results from the internal validation and test when the models are applied to the external test set