Lack of sufficiently strong informative features limits the potential of gene expression analysis as predictive tool for many clinical classification problems

Table 2 Fold difference in the expression values of informative probe sets for 6 different prediction problems in different data sets.

Data set	IBC versus Non-IBC				p53 mutation versus normal				PIK3 mutation versus normal
Receptor status	ER-negative		ER-positive		ER-negative		ER-positive		ER-negative		ER-positive
Phenotype	IBC	Non-IBC	IBC	Non-IBC	mutation	normal	mutation	normal	mutation	normal	mutation	normal
Pts #	19 vs 27		6 vs 31		44 vs 11		14 vs 31		8 vs 49		15 vs 57
Feature #	10	100	10	100	10	100	10	100	10	100	10	100
FDR q value	1.000	1.000	0.280	0.500	0.14	0.33	0.49	0.6	0.08	0.09	0.24	0.55
Fold difference *
< 0.5	0	20	0	2	1	9	2	29	0	0	0	17
≥0.5 - < 1.0	4	55	1	41	1	37	5	47	1	23	6	50
≥1.0 - < 1.2	2	11	1	10	1	11	1	11	3	18	1	17
≥1.2 - < 1.5	3	12	4	18	2	15	1	9	4	21	2	10
≥1.5 - < 2.0	0	0	2	20	2	22	1	4	1	23	0	5
≥2.0 - < 3.0	0	1	2	9	3	6	0	0	1	14	1	1
≥3.0 - < 4.0	1	1	0	0	0	0	0	0	0	1	0	0
≥4.0	0	0	0	0	0	0	0	0	0	0	0	0
Data set	WANG	TRANSBIG	Mainz
Receptor status	ER-negative	ER-positive	ER-negative	ER-positive	ER-negative	ER-positive
Recurrence	Yes	No	Yes	No	Yes	No	Yes	No	Yes	No	Yes	No
Pts #	29 vs 50	29 vs 50	18 vs 45	23 vs 112	11 vs 20	30 vs 139
Feature #	10	100	10	100	10	100	10	100	10	100	10	100
FDR q value	0.2	0.26	0.004	0.02	0.820	0.91	0.010	0.03	0.47	0.58	0.01	0.03
Fold difference *
< 0.5	2	35	3	68	2	26	1	26	1	13	3	59
≥0.5 - < 1.0	6	51	7	30	4	47	5	52	5	44	5	26
≥1.0 - < 1.2	0	7	0	2	3	21	2	12	0	5	1	3
≥1.2 - < 1.5	0	5	0	0	1	6	1	8	1	16	0	0
≥1.5 - < 2.0	1	1	0	0	0	0	1	2	2	17	1	2
≥2.0 - < 3.0	1	1	0	0	0	0	0	0	0	3	0	0
≥3.0 - < 4.0	0	0	0	0	0	0	0	0	1	2	0	0
≥4.0	0	0	0	0	0	0	0	0	0	0	0	0

* Log₂ Difference = abs(mean log₂ intensity for group1 - mean log₂ intensity for group 2) where "abs" is absolute value.
The numbers of probe sets with a given level of differential expression are shown for the 6 comparisons. Analyses were performed separately for Estrogen Receptor (ER)-positive and ER-negative cancers. IBC = inflammatory breast cancer, PI3K = Phosphatidylinositol-3 kinase, FDR = false discovery rate. The WANG, TRANSBIG, Mainz data sets correspond to references 2, 3 and 28. Probe sets with mean log₂ transformed expression difference > 1 between comparisons groups are highlighted in bold.

ISSN: 1471-2105