Table 7 Comparison of DRaW with IRNMF [8], AutoDTI++ [27], and DNILMF [66] on benchmark datasets [28]
From: DRaW: prediction of COVID-19 antivirals by deep learning—an objection on using matrix factorization
Dataset | Approach | AUC-ROC | AUPR |
|---|---|---|---|
Enzyme | IRNMF | 0.855 | 0.069 |
AutoDTI++ (\(S_{p}\)) | 0.90 | 0.82 | |
AutoDTI++ (\(S_{d}\)) | 0.50 | 0.33 | |
AutoDTI++ (\(S_{t}\)) | 0.84 | 0.77 | |
DNILMF | 0.981 | 0.727 | |
DRaW | 0.983 | 0.875 | |
Ion Channel | IRNMF | 0.817 | 0.144 |
AutoDTI++ (\(S_{p}\)) | 0.91 | 0.90 | |
AutoDTI++ (\(S_{d}\)) | 0.49 | 0.50 | |
AutoDTI++ (\(S_{t}\)) | 0.86 | 0.86 | |
DNILMF | 0.982 | 0.831 | |
DRaW | 0.983 | 0.886 | |
GPCR | IRNMF | 0.707 | 0.131 |
AutoDTI++ (\(S_{p}\)) | 0.86 | 0.85 | |
AutoDTI++ (\(S_{d}\)) | 0.47 | 0.47 | |
AutoDTI++ (\(S_{t}\)) | 0.85 | 0.83 | |
DNILMF | 0.954 | 0.648 | |
DRaW | 0.955 | 0.704 | |
Nuclear Receptor | IRNMF | 0.795 | 0.117 |
AutoDTI++ (\(S_{p}\)) | 0.87 | 0.84 | |
AutoDTI++ (\(S_{d}\)) | 0.60 | 0.62 | |
AutoDTI++ (\(S_{t}\)) | 0.87 | 0.84 | |
DNILMF | 0.919 | 0.626 | |
DRaW | 0.954 | 0.883 |