Table 2. Mean AUROC scores for the VDJdb data using leave-one-subject-out cross-validation.

TCRGP models and TCRdist models were trained using TCRβ with either only CDR3β or all CDRβs. RF models and DeepTCR models were trained using only CDR3β or CDR3β with Vβ-gene, from which the other CDRβs can be derived from.

	Method	TCRGP		TCRdist		RF		DeepTCR
	CDRs	3	all	3	all	3	3, V-gene	3	3, V-gene
	mean	0.806	0.863	0.743	0.783	NaN	0.796	0.776	0.777
CMV	pp65123-131	0.852	0.792	0.804	0.761	0.820	0.790	0.834	0.775
CMV	pp65417-426	0.912	0.968	0.829	0.874	0.930	0.930	0.890	0.860
CMV	pp65495-503	0.892	0.904	0.793	0.786	0.900	0.890	0.860	0.874
EBV	BMLF1280-288	0.926	0.947	0.863	0.872	0.930	0.910	0.897	0.907
EBV	BZLF1190-197	0.887	0.899	0.764	0.857	0.840	0.870	0.807	0.782
EBV	BRLF1109-117	0.682	0.756	0.729	0.762	0.770	0.780	0.670	0.564
IAV	M158-66	0.881	0.933	0.832	0.852	0.910	0.890	0.868	0.893
IAV	HA306-318	0.706	0.748	0.753	0.790	NaN	0.550	0.685	0.602
HCV	NS31073-1081	0.695	0.783	0.492	0.600	0.660	0.730	0.625	0.626
HCV	NS31406-1415	0.678	0.766	0.698	0.676	0.560	0.560	0.631	0.704
HCV	NS31436-1445	0.819	0.945	0.746	0.799	0.850	0.850	0.824	0.849
HSV-2	VP2249-57	0.801	0.814	0.721	0.713	0.850	0.860	0.858	0.823
YFV	NS4B214-222	0.825	0.823	0.692	0.695	0.890	0.870	0.649	0.664
DENV1	NS3133-142	0.864	0.914	0.878	0.920	0.900	0.920	0.906	0.876
DENV3-4	NS3133-142	0.734	0.859	0.842	0.893	0.530	0.710	0.850	0.843
HIV-1	p2430-40	0.894	0.886	0.836	0.808	0.880	0.880	0.818	0.858
HIV-1	p2448-56	0.817	0.946	0.716	0.800	0.720	0.810	0.624	0.794
HIV-1	p24128-135	0.770	0.740	0.541	0.594	0.720	0.720	0.719	0.771
HIV-1	p24131-140	0.838	0.859	0.701	0.724	0.760	0.790	0.740	0.656
HIV-1	p24161-180	0.750	0.992	0.769	0.975	0.780	0.890	0.846	0.844
HIV-1	p24223-231	0.702	0.894	0.686	0.819	NaN	0.520	0.651	0.703
HIV-1	Nef90-97	0.798	0.818	0.667	0.657	0.780	0.800	0.809	0.830