Table 1 Input feature vector composition from ablation experiments
Model | Input | Graph |
|---|---|---|
Model 1 | ESM-1b attention map (last layer only)  + one-hot encoding | Protein graph + Complete graph |
Model 2 | ESM-1b attention map (last layer only)  + one-hot encoding + ESM-1B representation | Only protein graph |
Model 3 | ESM-1b attention map (last layer only)  + one-hot encoding + ESM-1B representation | Only complete graph |
PCP-GC-LM | ESM-1b attention map (last layer only)  + one-hot encoding + ESM-1B representation | Protein graph + Complete graph |