Table 2.

Comparison of performance of MHCAttnNet with the state-of-the-art methods for class I MHC alleles

Model	AUC-	AUC-	PCC	PPV	F1-	Sensitivity
	ROC	PRC			score
NetMHC4.0	0.8237	0.9048	0.5738	0.9486	0.8531	0.7757
MHCflurry	0.8234	0.9074	0.5624	0.9695	0.8230	0.7149
PUFFIN	0.8185	0.9129	0.5398	0.9668	0.8264	0.7215
MHCAttnNet (no attention)	0.8822	0.9380	0.7463	0.9643	0.9402	0.9303
MHCAttnNet	0.8893	0.9418	0.7570	0.9683	0.9422	0.9304

Note: The best performances are indicated in bold.