Table 1 Performance of PPI-HotspotID,10, PPI-HotspotPLM,1280, PPI-HotspotPLM,300, and PPI-HotspotPLM,300-random on the same 70% split datasets
From: Using protein language models for protein interaction hot spot prediction with limited data
Method | PPI-HotspotID,10 | PPI-HotspotPLM,1280 | PPI-HotspotPLM,300 | PPI-HotspotPLM,300-random |
|---|---|---|---|---|
Architecture | Ensemble | Ensemble + ESM2 | Ensemble + ESM2 | Ensemble + ESM2 |
# of features | 10 | 1280 | 300 | 300 |
Input | Structure | Sequence | Sequence | Sequence |
Validation F1-score | 0.71 ± 0.002 | 0.69 ± 0.018 | 0.71 ± 0.02 | 0.70 ± 0.016 |