Table 2 Performance of different transfer strategies for plant secondary metabolic pathway prediction
From: Prediction of plant secondary metabolic pathways using deep transfer learning
Strategy | Accuracy (%) | Precision (%) | Recall (%) | F1_score (%) |
|---|---|---|---|---|
Entire model frozen | 97.04 ± 0.18 | 80.23 ± 0.99 | 70.90 ± 2.17 | 75.26 ± 1.44 |
Block1 frozen | 98.16 ± 0.20 | 86.99 ± 2.29 | 83.49 ± 1.10 | 85.20 ± 1.53 |
Block2 frozen | 98.30 ± 0.15 | 88.44 ± 1.59 | 84.36 ± 1.17 | 86.34 ± 1.07 |
Block1 + Block2 frozen | 98.15 ± 0.19 | 86.99 ± 2.26 | 83.45 ± 0.94 | 85.17 ± 1.40 |
No module frozen | 98.29 ± 0.15 | 88.23 ± 1.55 | 84.32 ± 1.31 | 86.22 ± 1.11 |