Table 2.
Performance comparison between the proposed comprehensive ensemble and the individual models on 19 bioassay datasets
| BioAssay | PubChem fingerprint | ECFP fingerprint | MACCS fingerprint | SMILES | comprehensive | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| RF | SVM | GBM | NN | RF | SVM | GBM | NN | RF | SVM | GBM | NN | NN | ensemble | |
| 1851_1a2 | 0.921 | 0.896 | 0.900 | 0.921 | 0.919 | 0.906 | 0.882 | 0.920 | 0.912 | 0.879 | 0.894 | 0.912 | 0.922 | 0.934 |
| 1851_2c19 | 0.871 | 0.852 | 0.848 | 0.872 | 0.882 | 0.871 | 0.854 | 0.880 | 0.874 | 0.842 | 0.850 | 0.885 | 0.875 | 0.900 |
| 1851_2c9 | 0.871 | 0.857 | 0.851 | 0.873 | 0.880 | 0.866 | 0.843 | 0.880 | 0.858 | 0.828 | 0.840 | 0.870 | 0.877 | 0.898 |
| 1851_2d6 | 0.858 | 0.847 | 0.832 | 0.869 | 0.867 | 0.850 | 0.833 | 0.856 | 0.854 | 0.816 | 0.830 | 0.852 | 0.846 | 0.884 |
| 1851_3a4 | 0.877 | 0.868 | 0.865 | 0.887 | 0.891 | 0.887 | 0.855 | 0.895 | 0.867 | 0.832 | 0.851 | 0.875 | 0.891 | 0.914 |
| 1915 | 0.754 | 0.692 | 0.709 | 0.722 | 0.731 | 0.700 | 0.700 | 0.712 | 0.758 | 0.716 | 0.736 | 0.741 | 0.701 | 0.755 |
| 2358 | 0.787 | 0.705 | 0.736 | 0.770 | 0.780 | 0.767 | 0.722 | 0.761 | 0.774 | 0.731 | 0.763 | 0.775 | 0.697 | 0.803 |
| 463213 | 0.673 | 0.639 | 0.652 | 0.651 | 0.685 | 0.652 | 0.644 | 0.661 | 0.668 | 0.642 | 0.655 | 0.651 | 0.636 | 0.689 |
| 463215 | 0.620 | 0.576 | 0.592 | 0.604 | 0.617 | 0.585 | 0.598 | 0.595 | 0.629 | 0.600 | 0.630 | 0.625 | 0.587 | 0.627 |
| 488912 | 0.679 | 0.643 | 0.634 | 0.668 | 0.693 | 0.654 | 0.668 | 0.675 | 0.667 | 0.634 | 0.650 | 0.673 | 0.644 | 0.698 |
| 488915 | 0.718 | 0.686 | 0.679 | 0.713 | 0.731 | 0.693 | 0.680 | 0.708 | 0.692 | 0.659 | 0.680 | 0.693 | 0.679 | 0.735 |
| 488917 | 0.808 | 0.777 | 0.759 | 0.805 | 0.814 | 0.788 | 0.760 | 0.799 | 0.788 | 0.726 | 0.752 | 0.786 | 0.780 | 0.834 |
| 488918 | 0.762 | 0.745 | 0.735 | 0.778 | 0.778 | 0.766 | 0.729 | 0.767 | 0.737 | 0.690 | 0.708 | 0.742 | 0.746 | 0.799 |
| 492992 | 0.829 | 0.784 | 0.783 | 0.800 | 0.849 | 0.807 | 0.802 | 0.822 | 0.825 | 0.726 | 0.759 | 0.790 | 0.802 | 0.845 |
| 504607 | 0.694 | 0.678 | 0.692 | 0.686 | 0.690 | 0.668 | 0.673 | 0.656 | 0.676 | 0.640 | 0.662 | 0.655 | 0.649 | 0.721 |
| 624504 | 0.884 | 0.850 | 0.857 | 0.867 | 0.884 | 0.858 | 0.858 | 0.861 | 0.872 | 0.832 | 0.862 | 0.876 | 0.868 | 0.897 |
| 651739 | 0.791 | 0.770 | 0.773 | 0.781 | 0.802 | 0.782 | 0.771 | 0.788 | 0.779 | 0.729 | 0.759 | 0.754 | 0.792 | 0.804 |
| 651744 | 0.884 | 0.862 | 0.872 | 0.885 | 0.889 | 0.883 | 0.875 | 0.896 | 0.869 | 0.829 | 0.843 | 0.853 | 0.899 | 0.901 |
| 652065 | 0.800 | 0.752 | 0.782 | 0.780 | 0.785 | 0.775 | 0.758 | 0.774 | 0.776 | 0.736 | 0.759 | 0.772 | 0.763 | 0.826 |
| average | 0.794 | 0.762 | 0.766 | 0.786 | 0.798 | 0.777 | 0.763 | 0.784 | 0.783 | 0.741 | 0.762 | 0.778 | 0.771 | 0.814 |
Each value shows the averaged AUC from twenty repeated experiments on the test set (bold: top 3 AUC on each dataset), and the last row shows the averaged AUC calculated from 19 AUC results