Table 6.

Probability of Predicting Seroreactive Antigens by Enriching Features^a) in Proteins Ranked by Naïve Bayes Classification

Top ranked antigens [%]	Antigens on chip	Seroreactive antigens	COGs			Computational prediction			MS Protein expression			Sum of all ten features
			Predicted	Hits	Rate	Predicted	Hits	Rate	Predicted	Hits	Rate	Predicted	Hits	Rate
1	30	7	15	6	0.40	30	7	0.23	27	5	0.19	30	7	0.23
2	60	13	30	10	0.33	60	13	0.22	46	9	0.20	60	13	0.22
5	152	33	53	15	0.28	151	33	0.22	76	16	0.21	152	33	0.22
10	304	54	97	24	0.25	270	47	0.17	129	25	0.19	304	54	0.18
20	617	77	136	27	0.20	440	59	0.13	289	36	0.12	617	77	0.12
25	761	88	176	34	0.19	488	59	0.12	341	40	0.12	753	88	0.12
30	913	96	237	36	0.15	544	64	0.12	344	41	0.12	866	95	0.11
40	1218	99	286	36	0.13	578	64	0.11	349	41	0.12	952	95	0.10
50	1523	107	302	36	0.12	633	65	0.10	353	41	0.12	1027	96	0.09
60	1827	112	304	36	0.12	668	67	0.10	355	41	0.12	1066	98	0.09
70	2132	115	318	36	0.11	670	67	0.10	356	41	0.12	1083	98	0.09
75	2284	115	320	36	0.11	670	67	0.10	356	41	0.12	1085	98	0.09
80	2436	118	321	36	0.11	686	67	0.10	356	41	0.12	1102	98	0.09
90	2732	122	329	37	0.11	691	67	0.10	356	41	0.12	1115	99	0.09
100	3046	122	338	37	0.11	696	67	0.10	356	41	0.12	1128	99	0.09

^a)The enriching features are classified into three categories: i) COGs (COG U, M, N, and O) for proteins assigned to one or more of these four COGs, ii) computational predictions with enriching features for proteins predicted to have one or more of the following features, i.e., TMHMM = 1, SignalP > 0.7, pSort outer membrane, pSort periplasmic, or pI < 5, and iii) protein expression as detected by mass spectrometry. For further details on the enriching features, cf. Table 1.