Table 2.
Sample sizes of the P2APP dataset, and the training and testing data used for the challenge
| Drug-indication pairs | Clinical trials | Unique drugs | Unique indications | Unique clinical trials | |
|---|---|---|---|---|---|
| Phase 2 to approval (P2APP) | |||||
| Success | 796 | 2,435 | 614 | 182 | 2,209 |
| Failure | 6,105 | 13,203 | 3,313 | 283 | 10,722 |
| Total | 6,901 | 15,638 | 3,726 | 291 | 12,680 |
| Training data | |||||
| Success | 610 | 1,852 | 468 | 169 | 1,666 |
| Failure | 4,293 | 6,839 | 2,537 | 264 | 5,845 |
| Total | 4,903 | 8,691 | 2,872 | 272 | 7,451 |
| Testing data | |||||
| Success | 186 | 583 | 160 | 93 | 557 |
| Failure | 1,812 | 6,364 | 1,096 | 218 | 5,065 |
| Total | 1,998 | 6,947 | 1,229 | 229 | 5,561 |
Note that the number of unique drugs, indications, and clinical trials are not necessarily additive across rows since drugs, indications, and trials have relationships that are surjective and non-injective: different drugs may target the same indication, and some trials may involve multiple drug-indication pairs. See also Note S1.