Table 1.
Summary of the training data
| Dataset | Source | Positives | Negatives | #HLAs | #cell lines | Data type |
|---|---|---|---|---|---|---|
| A | NetMHCpan | 42,020 | 128,550 | 112 | – | SA (BA) |
| 329,239 | 7,672,715 | 130 | 50 | SA and MA (EL) | ||
| B | EDGE | 105,672 | 1,344,404 | – | 69 | MA (EL) |
| C | HLAthena | 182,703 | 3,844,654 | 95 | – | SA (EL) |
| D | Trolle | 11,858 | 150,009 | 5 | – | SA (EL) |
SA refers to single allele, MA to multi-allele, BA to binding affinity, and EL to eluted ligand datasets. For BA data, the classification of positives and negatives was conducted using a threshold of 500 nM (Karosiene et al., 2013).