A comparison on effects of normalisations in the detection of differentially expressed genes

BMC Bioinformatics

Table 1 Area Under the Curve (AUC) of specificity and sensitivity of SAM test after the normalisations, for Albers' model with increasing percentage of background level with and without replacing negative values.

	10% bg		10% bg replaced		50% bg		50% bg replaced		150% bg		150% bg replaced
normalization	AUC	rank	AUC	rank	AUC	rank	AUC	rank	AUC	rank	AUC	rank
Raw	0.9	1	0.9	1	0.79	1	0.82	1	0.55	1	0.68	2
Global	0.91	2	0.91	2	0.84	2	0.86	2	0.66	2	0.74	4
GLOG	0.93	3	0.93	3	0.9	8	0.89	3	0.78	10	0.58	1
Lowess	0.94	6	0.93	3	0.86	3	0.9	6	0.69	6	0.73	3
P-Lowess	0.94	6	0.93	3	0.87	4	0.91	9	0.68	5	0.78	10
NeuralNet	0.93	3	0.93	3	0.87	4	0.91	9	0.69	6	0.76	5
OLIN	0.94	6	0.93	3	0.87	4	0.89	3	0.67	3	0.76	5
OSLIN	0.94	6	0.93	3	0.87	4	0.89	3	0.67	3	0.77	7
qsplineR	N0.94	6	0.93	3	0.92	9	0.9	6	0.72	8	0.77	7
qsplineG	0.93	3	0.93	3	0.92	9	0.9	6	0.72	8	0.77	7

For each simulated scenario is also reported the ranking of the normalisations according to the AUC: the bigger the rank, the better the normalisation.

ISSN: 1471-2105