Predicting MHC class I epitopes in large datasets

BMC Bioinformatics

Table 1 Prediction accuracies for the full dataset

	Allele Name	Binders	Non-Binders	Total	DynaPredPOS	NetMHC	SVMHC	YKW
		*(n)*	*(n)*	*(n)*	*AUC*	*AUC*	*AUC*	*AUC*
1	A*0101	163	1316	1479	0.93	0.98	0.95	0.94
2	A*0201	1544	1929	3473	0.93	0.96	0.92	0.91
3	A*0202	723	697	1420	0.88	0.93	0.85	0.85
4	A*0203	732	685	1417	0.88	0.95	0.86	0.84
5	A*0206	633	782	1415	0.88	0.95	0.87	0.86
6	A*0301	637	1618	2255	0.89	0.96	0.88	0.80
7	A*1101	816	1279	2095	0.92	0.96	0.90	0.91
8	A*2402	202	464	666	0.80	0.85	0.78	0.81
9	A*2601	69	885	954	0.84	0.93	0.83	0.84
10	A*3101	510	1480	1990	0.89	0.95	0.89	0.88
11	A*3301	203	994	1197	0.88	0.96	0.89	0.88
12	A*6801	578	620	1198	0.85	0.92	0.82	0.81
13	A*6802	439	980	1419	0.84	0.93	0.85	0.83
14	B*0702	238	1110	1348	0.94	0.98	0.94	0.93
15	B*0801	23	687	710	0.82	0.99	0.78	0.79
16	B*1501	182	836	1018	0.86	0.97	0.89	0.90
17	B*2705	81	917	998	0.93	0.97	0.90	0.94
18	B*3501	273	578	851	0.83	0.93	0.84	0.85
19	B*4001	94	1112	1206	0.90	0.97	0.93	0.91
20	B*4402	76	136	212	0.77	0.84	0.75	0.76
21	B*4403	71	142	213	0.68	0.81	0.65	0.70
22	B*5101	108	249	357	0.82	0.93	0.80	0.81
23	B*5301	127	228	355	0.86	0.95	0.85	0.87
24	B*5801	78	893	971	0.90	0.99	0.93	0.93
	Average AUC				0.86	0.94	0.86	0.86

Alleles included in the study with the number of binders and non-binders available; all available binders and non-binders were included in the analysis irrespective of whether quantitative laboratory test data was available or not (Dataset F). Only unique peptide sequences were included in the counts; all peptides with more than one entry for a particular allele in IEDB were counted once only. The overall performance of the four prediction models on different alleles is shown; AUC = area under the curve (ROC analysis). The average AUC for each method is included at the bottom of each column.

ISSN: 1471-2105