Table 2.
Ensemble inference performance results comparison.
| Union (7) | CBR | probCBR | Rephetio | TransE | DistMult | ComplEx | RotatE | |
|---|---|---|---|---|---|---|---|---|
|
| ||||||||
| Positive | 80% (20) | 28% (7) | 48% (12) | 48% (12) | 72% (18) | 64% (16) | 68% (17) | 48% (12) |
| Neutral | 16% (4) | 60% (15) | 48% (12) | 32% (8) | 24% (6) | 36% (9) | 28% (7) | 48% (12) |
| Negative | 4% (1) | 12% (3) | 4% (1) | 20% (5) | 4% (1) | 0% (0) | 4% (1) | 4% (1) |
Through literature curation, the top predicted disease for each drug for each given algorithm was categorized into three groups by the effect the drug has on its predicted indication; the groups are positive, neutral, and negative effects. An indication was categorized as positive if a drug improved the predicted disease outcome, negative if a drug exacerbated the predicted disease, or neutral if a predicted disease was neither affected nor associated with a drug.