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

Intensity-dependent normalization is important for both high-density oligonucleotide array and cDNA array data. Both the regression and spline-based methods described here performed better than existing linear methods when assessed on the variability of replicate arrays. Dye-swap normalization was less effective at Cy(3)-Cy(5) normalization than either regression or spline-based methods alone.