Table 1.
Prediction performance of selected representative servers at two scenarios: a) thresholds that correctly predict ~80% of T-cell epitopes; b) thresholds that correctly predict ~50% of T-cell epitopes.
| Server | Threshold | TP | FN | TN | FP | TP (tumor epitopes) |
| a) 80% prediction | ||||||
| IEDB_ARB | 100 | 10 | 5 | 56 | 32 | 17 (81%) |
| IEDB_ CON | 11 | 10 | 5 | 70 | 18 | 17 (81%) |
| MHCPRED | 15 | 10 | 5 | 36 | 52 | 17 (81%) |
| MULTI_ANN | 3 | 15 | 0 | 38 | 50 | 17 (81%) |
| MULTI_SVM | 6.4 | 13 | 2 | 59 | 29 | 17 (81%) |
| NETMHCII | 4000 | 12 | 3 | 51 | 37 | 17 (81%) |
| NETMHCIIPAN | 440 | 12 | 3 | 56 | 32 | 17 (81%) |
| PROPRED | -1.5 | 14 | 1 | 54 | 34 | 17 (81%) |
| RANKPEP | 0.85 | 12 | 3 | 41 | 47 | 17 (81%) |
| b) 50% prediction | ||||||
| IEDB_ARB | 3 | 7 | 8 | 80 | 8 | 10 (48%) |
| IEDB_ CON | 4 | 7 | 8 | 86 | 2 | 10 (48%) |
| MHCPRED | 3 | 4 | 11 | 76 | 12 | 10 (48%) |
| MULTI_ANN | 5 | 5 | 10 | 66 | 22 | 11 (52%) |
| MULTI_SVM | 7.1 | 6 | 9 | 82 | 6 | 10 (48%) |
| NETMHCII | 400 | 7 | 8 | 78 | 10 | 10 (48%) |
| NETMHCIIPAN | 60 | 7 | 8 | 87 | 1 | 10 (48%) |
| PROPRED | 0 | 10 | 5 | 81 | 7 | 11 (52%) |
| RANKPEP | 7 | 10 | 5 | 69 | 19 | 11 (52%) |