Table 4.
Model performances on the FDA ER-EDKB test set
| Algorithm | Training set | Variable type | Pearson correlation |
|---|---|---|---|
| RF | ALL+Xeno | @TOME+LD | 0.748 |
| RF | ALL+Xeno | @TOME+LD+MACCS | 0.740 |
| RF | ALL | @TOME+LD | 0.663 |
| RF | ALL | @TOME+LD+MACCS | 0.648 |
| RF | BDB+Xeno | @TOME+LD | 0.712 |
| RF | BDB+Xeno | @TOME+LD+MACCS | 0.688 |
| RF | BDB | @TOME+LD | 0.584 |
| RF | BDB | @TOME+LD+MACCS | 0.542 |
| RF | BDB+Xeno | MACCS only | 0.487 |
Note: The presented models employ all the RF algorithm and differ in training set composition concerning used molecules and in type of variables used. @TOME+LD = docking evaluation variables from the @TOME server + ligand descriptors calculated with CDK.