Table 3.

Error rates for classifiers trained on one data set and tested on other public data sets

	BL error ratea				DLBCL error ratea
Normalization	Z-score	Rank	XPN	DWD	Z-score	Rank	XPN	DWD
Train GSE4732_p1: test on other data sets below
GSE4475 (strict)b	0.09	0.09	0.09	0.09	0.017	0.017	0.006	0
GSE4732_p2	0.182	0.212	0.152	0.152	0	0	0	0
GSE10172 (strict)b	0.231	0.308	0.385	0.308	0	0	0	0
GSE26673 eBL	0.615	0.692	0.846	0.384
GSE26673 and GSE17189 HIV-related	0.833	1	1	0.667	0	0	0	0
Train GSE4475 strict BL definition: test on other data sets below
GSE4732_p1	0.04	0.04	0.04	0.04	0.012	0.008	0.012	0.012
GSE4732_p2	0.303	0.333	0.273	0.273	0	0	0	0
GSE10172 (strict)	0.154	0.154	0.308	0.154	0	0	0	0
GSE26673 eBL	0.615	0.538	0.769	0.538
GSE26673 and GSE17189 HIV-related	0.833	0.833	1	0.833	0	0	0	0
Train GSE4475 wide BL definition: test on other data sets below
GSE4732_p1	0.02	0.02	0.02	0.02	0.04	0.05	0.06	0.07
GSE4732_p2	0.06	0.03	0.03	0.03	0.015	0.015	0.015	0.015
GSE10172 (strict)	0.078	0.078	0	0.078	0.043	0.043	0	0.043
GSE26673 eBL	0.154	0.154	0.308	0.154
GSE26673 and GSE17189 HIV-related	0.5	0.333	0.833	0.5	0	0	0	0

^aError rate is (1 − Recall) value for the indicated class [Recall = True positives/(True positives + False negatives)]

^bThe sample in this data set is assigned to mBL, intermediate, non-mBL categories; here we set the strict BL definition as the standard which put intermediate and non-mBL together as the DLBCL class. eBL endemic BL, mBL molecular BL