A new non-linear normalization method for reducing variability in DNA microarray experiments

Table 1 Variance of replicates

	Pre	Scaled	Lowess	Lowess-tip	Qspline	Qspline-tip	Spatial
T cells	0.140	0.062	0.062	-	0.057	-	0.056
HIV	0.097	0.042	0.032	-	0.031	-	0.030
glnA.Cy3	0.449	0.276	0.394	0.390	0.287	0.251	0.264
glnA.Cy5	0.643	0.634	0.553	0.539	0.389	0.326	0.366
gnrA.Cy3	0.400	0.316	0.372	0.366	0.326	0.315	0.273
tnrA.Cy5	0.466	0.473	0.393	0.384	0.353	0.318	0.338
T cells	-	55.4	55.9	-	59.5	-	60.1
HIV	-	56.5	66.9	-	67.9	-	68.8
glnA.Cy3	-	38.6	12.2	13.1	36.1	44.1	41.2
glnA.Cy5	-	1.4	14.0	16.2	39.4	49.4	43.1
tnrA.Cy3	-	21.2	7.2	8.7	18.6	21.4	31.8
tnrA.Cy5	-	0.0	15.6	17.5	24.1	31.6	27.4

The top half of the table shows average log-signal variances of oligonucleotide array replicates (T-cell control, HIV-infected) and cDNA array replicates (glnA.Cy3 and tnrA.Cy3 controls, glnA.Cy5 and tnrA.Cy5 mutants) before normalization (pre), after log-signal scaling (scaled), global lowess (lowess), scaled print-tip lowess (lowess-tip), global and tip scaled qspline (qspline, qspline-tip) and spatial gradient normalization (spatial). The percent decrease relative to prenormalized variance is also listed for each method in the lower half of the table.

ISSN: 1474-760X