Table 2 Metrics calculated for the final selections of high-scoring molecules according to the LogP predictor (Score > 0.5)
From: Human-in-the-loop active learning for goal-oriented molecule generation
Metric | No Feedback | Feedback (\(T = 10\)) | |||
|---|---|---|---|---|---|
EPIG (\(\sigma _{\epsilon } = 5.0\)) | Uncertainty (\(\sigma _{\epsilon } = 5.0\)) | Greedy (\(\sigma _{\epsilon } = 5.0\)) | Random (\(\sigma _{\epsilon } = 5.0\)) | ||
Number of molecules | 124.44 ± 1.34 | 125.38 ± 1.58 | 126.22 ± 1.40 | 125.89 ± 1.52 | 124.78 ± 1.62 |
MAE Oracle-Pred. \(\downarrow\) | 2.15 ± 0.24 | 1.35 ± 0.16 ** | 1.29 ± 0.21 ** | 1.42 ± 0.19 ** | 1.91 ± 0.50 |
Internal Diversity \(\uparrow\) | 0.85 ± 0.01 | 0.84 ± 0.01 | 0.84 ± 0.01 * | 0.84 ± 0.01 | 0.85 ± 0.01 |
SA \(\downarrow\) | 3.12 ± 0.15 | 2.80 ± 0.12 ** | 2.78 ± 0.15 ** | 2.83 ± 0.06 ** | 3.04 ± 0.16 |
QED \(\uparrow\) | 0.52 ± 0.03 | 0.45 ± 0.03 ** | 0.44 ± 0.03 ** | 0.46 ± 0.04 ** | 0.51 ± 0.05 |
Novelty \(\uparrow\) | 1.0 ± 0.0 | 1.0 ± 0.0 | 1.0 ± 0.0 | 1.0 ± 0.0 | 1.0 ± 0.0 |
Uniqueness \(\uparrow\) | 1.0 ± 0.0 | 1.0 ± 0.0 | 1.0 ± 0.0 | 1.0 ± 0.0 | 1.0 ± 0.0 |
Frag Gen-Train \(\uparrow\) | 0.70 ± 0.05 | 0.85 ± 0.05 ** | 0.88 ± 0.03 ** | 0.88 ± 0.04 ** | 0.86 ± 0.04 ** |
SNN Gen-Train \(\uparrow\) | 0.23 ± 0.01 | 0.26 ± 0.01 ** | 0.27 ± 0.01 ** | 0.26 ± 0.01 ** | 0.25 ± 0.01 ** |
FCD Gen-Train \(\downarrow\) | 35.64 ± 1.31 | 30.07 ± 1.07 ** | 30.88 ± 1.69 ** | 31.81 ± 1.07 ** | 32.30 ± 1.62 ** |
Frag Gen-Queries \(\uparrow\) | - | 0.96 ± 0.02 | 0.95 ± 0.02 | 0.94 ± 0.02 | 0.92 ± 0.03 |
SNN Gen-Queries \(\uparrow\) | - | 0.27 ± 0.01 | 0.27 ± 0.02 | 0.26 ± 0.00 | 0.25 ± 0.01 |
FCD Gen-Queries \(\downarrow\) | - | 27.12 ± 1.03 | 27.45 ± 0.87 | 27.23 ± 1.07 | 26.47 ± 1.37 |