Table 3.
Percentages of valid, unique and ZINC15 database-matched SMILES strings generated by the model in ten per one mode.
Dataset 1 | Dataset 2 | Dataset 3 | Dataset 4 | Dataset 5 | Average | |
---|---|---|---|---|---|---|
Total number of generated SMILES strings (ten per target protein) | 1040 | 1120 | 1030 | 1220 | 1240 | 1130 |
Valid | 80.8 | 80.5 | 79.6 | 86.2 | 86.1 | 82.6 |
Unique | 88.4 | 78.4 | 71.7 | 85.0 | 85.0 | 81.7 |
Match with ZINC15 database (%) | 17.9 | 16.3 | 10.7 | 19.5 | 21.2 | 17.1 |