Table 2. Results of a Small Virtual Screen against Tb REL1 Used to Test the Novel Neural-Network Scoring Functiona.
top 10b | EFc | rankATPd | rankV1e | rankS5f | |
---|---|---|---|---|---|
Vina | 2/3 | 6.9 | 8 | 18 | 9 |
NN1 | 1/3 | 3.4 | 99 | 2 | 25 |
NN2 | 1/3 | 3.4 | 2 | 16 | 24 |
NN3 | 2/3 | 6.9 | 7 | 1 | 69 |
N (average24) | 2/3 | 6.9 | 8 | 5 | 27 |
Five scoring functions compared: AutoDock Vina score, predictions of the top three individual neural networks (NN1, NN2, and NN3, respectively), and average prediction of the top 24 networks (N).
Number of known inhibitors that ranked in the top 10 for each scoring function.
Enrichment factor when the top 10 ligands were considered.
Rank of the known inhibitor ATP.
Rank of the known inhibitor V1.
Rank of the known inhibitor S5.