Table 1.
The 5-fold cross-validation performance of MHCAttnNet with different hyper-parameters on class I MHC alleles
| Bi-LSTM | Number of layers | Number of layers | Context | AUC-PRC | AUC-ROC | F1- |
|---|---|---|---|---|---|---|
| hidden dimension | in peptide Bi-LSTM | in MHC Bi-LSTM | dimension | score | ||
| 64 | 3 | 3 | 16 | 0.9418 | 0.8893 | 0.9422 |
| 64 | 3 | 1 | 16 | 0.9418 | 0.8893 | 0.9416 |
| 64 | 3 | 3 | 32 | 0.9409 | 0.8874 | 0.9398 |
| 64 | 3 | 1 | 32 | 0.9416 | 0.8887 | 0.9404 |
Note: The best performance is indicated in bold. The hyper-parameter setting, corresponding to this, was used to run all the experiments.