Summary of molecular descriptors selected and training and test set correlations for MIC predictions
Prediction Model | Selected Descriptors | Training (r2) | Test Set (r2) |
|---|---|---|---|
| Pharmacophores | Hydrophobic feature and hydrogen bond acceptor | N.A. | N.A. |
| ChemTree (single- and 100-tree models) | Hydrogen bond acceptor and hydrophobic | 49/54 = 91% | 10/12 = 83.3% |
| ChemTree 100-tree models with Cerius2 descriptors | Hydrogen bond acceptor, hydrophobic, and radius of gyration | 49/54 = 91% | 10/12 = 83.3% |
| Cerius2 CSAR tree | Hydrogen bond acceptor, hydrogen bond donor, area, and dipole magnitude | 54/54 = 100% | 10/12 = 83.3% |
| Tree in R | Hydrogen bond acceptor, radius of gyration, sum of atomic polarization, and dipole magnitude | 52/54 = 96% | 11/12 = 91.6% |
| Linear model in R | Hydrogen bond acceptor | 43/54 = 80% | 11/12 = 91.6% |
N.A., not applicable.