Table 2.
Comparison of the performance of the different methods in the multi-target case
| Rewarding scheme | Dataset | Validity | Desirability | Uniqueness | Diversity | Purine ring | Furan ring | Benzene ring |
|---|---|---|---|---|---|---|---|---|
| LIGAND | 100.00% | 12.40% | 100.00% | 0.66 | 21.30% | 35.44% | 79.24% | |
| PF | DrugEx v1 | 98.28% | 43.27% | 88.96% | 0.71 | 17.37% | 41.05% | 80.95% |
| DrugEx v2 | 99.57% | 80.81% | 87.29% | 0.7 | 13.97% | 32.01% | 80.26% | |
| ORGANIC | 98.84% | 66.01% | 82.67% | 0.65 | 17.27% | 56.38% | 68.87% | |
| REINVENT | 99.54% | 57.43% | 98.84% | 0.77 | 0.64% | 40.38% | 92.05% | |
| WS | DrugEx v1 | 97.76% | 38.44% | 93.44% | 0.71 | 10.76% | 36.42% | 86.99% |
| DrugEx v2 | 99.80% | 97.45% | 89.08% | 0.49 | 3.63% | 21.06% | 96.18% | |
| ORGANIC | 99.08% | 61.10% | 77.65% | 0.68 | 9.08% | 70.99% | 83.91% | |
| REINVENT | 99.54% | 70.98% | 99.11% | 0.71 | 0.04% | 23.23% | 96.28% |
Shown are validity, desirability, uniqueness, and substructure distributions of SMILES generated by four different methods in the multi-target case with PF and WS rewarding schemes. For the validity, desirability and uniqueness, the highest values are bold, while for the distribution of substructures, the bold data are labeled as the closest to the values in the LIGAND set