Table 4 Comparison of adverse drug prediction models in the literature
From: Prediction of adverse drug reactions based on knowledge graph embedding
Study | Drug | Features | ADRs | Label Source | AUC all | AUC top10 |
|---|---|---|---|---|---|---|
Our work | 3632 | Drug targets, indications, and ADRs | 5589 | SIDER | 0.86 | 0.934 |
Luo et al.[12] | 655 | Molecular docking (600 proteins) | 1533 | SIDER | 0.84 | – |
LaBute et al.[24] | 560 | Molecular docking (409 proteins) | 85 (10 groups) | SIDER | 0.60–0.69 | – |
Bean et al.[13] | 524 | Drug targets, indications, and ADRs | 10 | SIDER/EHR | – | 0.92 |
Cao et al.[25] | 746 | Structures, gene expressions, and multiple evidences sources | 817 | SIDER | 0.57-0.88 | – |
Jamal et al.[26] | 928 | Chemical, biological, and phenotypic properties | 22 | SIDER | 0.48–1.00 | – |
Hu et al.[11] | 548 | DDI, PPI, drug target and treatment information, chemical structures, and side effects | 1318 | SIDER | 0.84 | – |
Dey et al.[10] | 1430 | Structure information and side effects | 1766 | SIDER | – | 0.919 |