Model selection and efficiency testing for normalization of cDNA microarray data

Table 1 Number of significant spot neighborhoods on the intensity scale for arrays of the experiments analyzed

Microarray	No normalization	Linear normalization	Global lowess	P-lowess	S-lowess	LIN	OLIN	OSLIN
SW 1	596	580	2	5	7	39	12	0
SW 2	1,090	1,080	0	0	0	52	0	0
SW 3	1138	988	0	0	0	0	0	0
SW 4	745	655	0	9	4	0	12	0
Apo 1	1,748	1,810	18	0	0	26	0	0
Apo 2	2,739	2,683	16	17	26	24	0	0
Apo 3	3,479	3,559	52	30	15	0	1	0
Apo 4	2,122	2,184	1	17	22	0	0	11
Apo 5	3,885	3,886	100	13	11	94	0	0
Apo 6	3,540	3,555	0	0	0	114	0	0
Apo 7	3,725	3,724	0	0	7	0	0	0
Apo 8	3,253	3,296	66	0	0	507	0	0
Apo 9	1,040	1,044	114	0	0	455	0	0
Apo 10	1,554	1,575	0	0	0	45	0	0
Apo 11	2,903	2,889	5	34	35	2	0	0
Apo 12	3,536	3,543	14	0	0	47	0	0
Apo 13	2,819	2,901	132	12	3	354	4	0
Apo 14	2,291	2,342	313	0	12	484	64	4
Apo 15	3,050	3,034	95	36	0	164	0	0
Apo 16	1,338	1,374	159	0	12	918	0	0

Spot neighborhoods are found significant if the median log ratio M has larger positive and negative value than expected from random order of spots along the intensity scale. The level of significance was defined by FDR = 0.01. The spot neighborhood was defined by a symmetrical window of 50 spots.

ISSN: 1474-760X