Table 2:
Composition of the datasets used in this paper. ReDocked2020 and CrossDocked2020 both have model-generated counterexample. CrossDocked Iteration 0 is the CrossDocked2020 set without any counterexamples added. ReDocked2020 and CrossDocked Only form a non-overlapping partition of CrossDocked2020 into redocked and cross-docked poses. Affinity Data refers to the percentage of poses with associated binding affinities from the PDBbind.
| Dataset | Pockets | Complexes | Poses | Ligands | Affinity Data % |
|---|---|---|---|---|---|
| PDBbind Core | – | 280 | 4,618 | 280 | 100 |
| PDBbind Refined | – | 3,805 | 66,953 | 2,972 | 100 |
| PDBbind General | – | 11,324 | 201,839 | 8,757 | 100 |
| ReDocked2020 | 2,916 | 18,369 | 786,960 | 13,780 | 32.7 |
| CrossDocked Iteration 0 | 2,922 | 18,450 | 10,691,929 | 13,839 | 39.9 |
| CrossDocked Iteration 1 | 2,922 | 18,450 | 19,182,423 | 13,839 | 41.3 |
| CrossDocked Only | 2,767 | 18,293 | 21,797,142 | 13,786 | 42.2 |
| CrossDocked2020 | 2,922 | 18,450 | 22,584,102 | 13,839 | 41.9 |