Table 3.

Tenfold cross validation performance on strict split

Model	ROC-AUC
K-NN (Baseline)	0.54 ± 0.03
AA CDR3	0.60 ± 0.04
AA full	0.59 ± 0.04
SMI CDR3	0.60 ± 0.06
SMI full	0.59 ± 0.06
Pretrained	0.56 ± 0.04
Pretrained + Aug.	0.59 ± 0.03
Pretrained semifrozen	0.58 ± 0.06
Pretrained semifrozen + Aug.	$0.62\pm 0.06$

Note: Mean and standard deviation of each model configuration on the VDJ + COVID-19 dataset. Best performance marked in bold.