Skip to main content
. 2021 Jul 12;37(Suppl 1):i237–i244. doi: 10.1093/bioinformatics/btab294

Table 4.

Comparison of TITAN with prior work

ROC-AUC
Model Seen epitopes Unseen epitopes
ImRex 0.61 0.50
K-NN (Baseline) 0.79 0.37
AA CDR3 0.83 0.69
AA full 0.81 0.64
SMI CDR3 0.85 0.72
SMI full 0.86 0.64
Pretrained 0.79 0.78
Pretrained + Aug. 0.83 0.65
Pretrained semifrozen 0.77 0.69
Pretrained semifrozen + Aug. 0.87 0.60

Note: All but the ImRex model (shaded in gray) are contributions of this work. Models were trained on identical data and tested on an independent test set to ensure a fair comparison. Best performance marked in bold.