Table 4 Performance of models 2, 3, 4 trained in schemes 3, 4, 5, 6 and tested in schemes 1, 2, 5, 6 on the PDBbind v2013 blind benchmark
From: Correcting the impact of docking pose generation error on binding affinity prediction
Model | Training | Test | RMSE | SD | Rp | Rs |
|---|---|---|---|---|---|---|
2 (MLR::Vina) | 3 | 1 | 1.70 | 1.69 | 0.521 | 0.511 |
3 (RF::Vina) | 3 | 1 | 1.60 | 1.58 | 0.602 | 0.575 |
4 (RF::VinaElem) | 3 | 1 | 1.48 | 1.48 | 0.666 | 0.643 |
2 (MLR::Vina) | 3 | 2 | 1.69 | 1.69 | 0.523 | 0.509 |
3 (RF::Vina) | 3 | 2 | 1.59 | 1.58 | 0.601 | 0.562 |
4 (RF::VinaElem) | 3 | 2 | 1.49 | 1.49 | 0.655 | 0.635 |
2 (MLR::Vina) | 4 | 1 | 1.88 | 1.80 | 0.413 | 0.415 |
3 (RF::Vina) | 4 | 1 | 1.72 | 1.71 | 0.499 | 0.477 |
4 (RF::VinaElem) | 4 | 1 | 1.57 | 1.57 | 0.610 | 0.589 |
2 (MLR::Vina) | 4 | 2 | 1.77 | 1.75 | 0.468 | 0.447 |
3 (RF::Vina) | 4 | 2 | 1.70 | 1.66 | 0.544 | 0.508 |
4 (RF::VinaElem) | 4 | 2 | 1.58 | 1.57 | 0.611 | 0.582 |
2 (MLR::Vina) | 5 | 5 | 1.65 | 1.65 | 0.550 | 0.526 |
3 (RF::Vina) | 5 | 5 | 1.58 | 1.58 | 0.603 | 0.578 |
4 (RF::VinaElem) | 5 | 5 | 1.49 | 1.50 | 0.653 | 0.633 |
2 (MLR::Vina) | 6 | 6 | 1.68 | 1.68 | 0.526 | 0.514 |
3 (RF::Vina) | 6 | 6 | 1.57 | 1.57 | 0.608 | 0.581 |
4 (RF::VinaElem) | 6 | 6 | 1.47 | 1.48 | 0.665 | 0.643 |