Table 2.
EmbedPVP variant prediction results for inductive versus transductive approach for the HP model using the clinical phenotypes and selected models based on the best-performing models using the transductive approach.
| H@1 | H@10 | H@30 | H@50 | ROCAUC | AUPR | ||
|---|---|---|---|---|---|---|---|
| EmbedPVP (TransE) | Inductive | 204 (0.1335) | 388 (0.2539) | 535 (0.3501) | 678 (0.4437) | 0.8918 | 0.0840 |
| Transductive | 218 (0.1427) | 415 (0.2716) | 589 (0.3855) | 710 (0.4647) | 0.9144 | 0.0908 | |
| EmbedPVP (TransD) | Inductive | 168 (0.1099) | 491 (0.3213) | 706 (0.4620) | 847 (0.5543) | 0.9631 | 0.0790 |
| Transductive | 482 (0.3154) | 846 (0.5537) | 1007 (0.6590) | 1056 (0.6911) | 0.9895 | 0.2507 | |
| EmbedPVP (DL2Vec) | Inductive | 202 (0.1322) | 409 (0.2677) | 558 (0.3652) | 613 (0.4012) | 0.9776 | 0.0868 |
| Transductive | 362 (0.2369) | 666 (0.4359) | 787 (0.5151) | 826 (0.5406) | 0.9867 | 0.1758 | |
| EmbedPVP (OWL2Vec*) | Inductive | 179 (0.1171) | 386 (0.2526) | 543 (0.3554) | 612 (0.4005) | 0.9776 | 0.0764 |
| Transductive | 409 (0.2677) | 685 (0.4483) | 783 (0.5124) | 842 (0.5510) | 0.9874 | 0.1987 |