Table 2 Performance comparison of five methods on four internal protein targets
From: Deep scaffold hopping with multimodal transformer neural networks
Metrics | Models | Â | |||
|---|---|---|---|---|---|
LBVS | MMPA | Seq2seq | G2G | DeepHop | |
Success Rate (%) | 34.6(16.4) | 33.9(13.2) | 29.1(16.6) | 33.4(6.6) | 65.1(12.7) |
Constraint Success (%) | 10.4(7.2) | 13.4(6.1) | 15.9(8.5) | 14.7(1.5) | 33.5(8.8) |
Improvement | − 0.94(0.61) | 0.31(0.15) | 0.77(0.47) | 0.88(0.65) | 0.81(0.32) |
Validity (%) | – | 97.3(0.04) | 29.2(0.11) | 99.7(0.03) | 93.9(0.02) |
Uniqueness (%) | – | 99.2(0.01) | 17.8(0.03) | 17.9(0.15) | 70.7(0.05) |
Novelty (%) | – | – | 99.1(0.03) | 100(0.00) | 99.5(0.01) |