Derivation of an amino acid similarity matrix for peptide:MHC binding and its application as a Bayesian prior

Table 2 Prediction performances as measured by AUC values of ROC curves.

MHC	Data Size	SMM	SMM^PMBEC	NetMHC
H-2_Db	303	0.909	0.901	0.933
H-2_Dd	85	0.813	0.837	0.925
H-2_Kb	223	0.811	0.833	0.850
H-2_Kd	176	0.928	0.931	0.939
H-2_Kk	164	0.772	0.793	0.790
H-2_Ld	102	0.932	0.942	0.977
HLA_A-0101	1157	0.977	0.977	0.982
HLA_A-0201	3089	0.946	0.946	0.957
HLA_A-0202	1447	0.898	0.899	0.900
HLA_A-0203	1443	0.916	0.916	0.921
HLA_A-0206	1437	0.913	0.916	0.927
HLA_A-0301	2094	0.927	0.928	0.937
HLA_A-1101	1985	0.938	0.939	0.951
HLA_A-2301	104	0.793	0.840	0.852
HLA_A-2402	197	0.803	0.801	0.825
HLA_A-2403	254	0.919	0.932	0.918
HLA_A-2601	672	0.916	0.924	0.956
HLA_A-2902	160	0.912	0.916	0.935
HLA_A-3001	669	0.935	0.941	0.947
HLA_A-3002	92	0.878	0.830	0.744
HLA_A-3101	1869	0.925	0.925	0.928
HLA_A-3301	1140	0.923	0.925	0.915
HLA_A-6801	1141	0.885	0.885	0.883
HLA_A-6802	1434	0.899	0.899	0.899
HLA_A-6901	833	0.867	0.880	0.880
HLA_B-0702	1262	0.959	0.960	0.965
HLA_B-0801	708	0.932	0.956	0.955
HLA_B-1501	978	0.937	0.940	0.941
HLA_B-1801	118	0.881	0.880	0.838
HLA_B-2705	969	0.936	0.941	0.938
HLA_B-3501	736	0.883	0.889	0.875
HLA_B-4002	118	0.832	0.843	0.754
HLA_B-4402	119	0.731	0.739	0.778
HLA_B-4403	119	0.757	0.753	0.763
HLA_B-4501	114	0.825	0.866	0.862
HLA_B-5101	244	0.876	0.895	0.886
HLA_B-5301	254	0.889	0.885	0.899
HLA_B-5401	255	0.923	0.935	0.903
HLA_B-5701	59	0.826	0.843	0.826
HLA_B-5801	988	0.942	0.945	0.961
Mamu_A-01	525	0.855	0.854	0.861
Mamu_A-02	283	0.765	0.783	0.809
Mamu_A-11	468	0.883	0.894	0.894
Mamu_B-01	205	0.949	0.956	0.967
Mamu_B-17	300	0.934	0.943	0.954
Patr_B-0101	132	0.964	0.975	0.969
average AUC		0.887	0.894	0.897
t-test\|SMM		NA	0.001	0.057
t-test\|NetMHC		0.057	0.470	NA

For each MHC-specific binding data set, 5-fold cross validation was carried out using one of the four methods listed. Area-Under-Curve (AUCs) from the 5-fold cross validations are presented here. Also shown are p-values of average AUC values calculated using Student's t-test (2-tail, paired) with respect to the SMM and NetMHC prediction methods.