Table 2 Performance comparison on property prediction of small molecules (regression tasks)
From: Force field-inspired molecular representation learning for property prediction
Metric | RMSE ↓ | MAE ↓ | ||
|---|---|---|---|---|
Dataset | ESOL | Lipophilicity | FreeSolv | QM9 |
SVM | 1.128 (0.081) | 0.785 (0.032) | 2.283 (0.324) | –a |
RF | 1.206 (0.034) | 0.859 (0.030) | 2.093 (0.566) | 14.584 (0.047) |
GATv2 | 0.578 (0.031) | 0.618 (0.014) | 1.017 (0.122) | 3.449 (0.146) |
GIN | 0.619 (0.044) | 0.756 (0.007) | 1.136 (0.235) | 4.972 (0.263) |
GCN | 0.778 (0.101) | 0.899 (0.035) | 1.582 (0.325) | 10.158 (0.236) |
DMPNN | 0.665 (0.052) | 0.596 (0.050) | 1.167 (0.150) | 3.101 (0.010) |
DimeNet | 0.730 (0.154) | 0.699 (0.096) | 0.890 (0.191) | 0.748 (0.065) |
FFiNet | 0.551 (0.030) | 0.579 (0.022) | 0.756 (0.138) | 1.803 (0.102) |