Table 3 Top-k accuracy\((\%)\) after re-ranking over R-SMILES on USPTO-50k
From: RetroRanker: leveraging reaction changes to improve retrosynthesis prediction through re-ranking
Models | Top-1 | Top-3 | Top-5 | Top-10 | Top-20 |
|---|---|---|---|---|---|
RetroSim | 37.3 | 54.7 | 63.3 | 74.1 | 82.0 |
RetroXpert | 50.4 | 61.1 | 62.3 | 63.4 | 63.9 |
GLN | 52.5 | 69.0 | 75.6 | 83.7 | 89.0 |
LocalRetro | 53.4 | 77.5 | 85.9 | 92.4 | - |
AT | 53.5 | – | 81.0 | 85.7 | - |
R-SMILES | 56.0 | 79.1 | 86.1 | 91.0 | 93.5 |
R-SMILES+RetroRanker\(^{*}\) | 56.0 | 79.6 (+0.5) | 86.6 (+0.5) | 91.8 (+0.8) | 93.8 (+0.3) |
R-SMILES+RetroRanker\(^{\dagger }\) | 56.0 | 79.5 (+0.4) | 86.9 (+0.8) | 91.5 (+0.5) | 94.0 (+0.5) |