By scanning the arrays, we obtained the raw hybridization intensities for each element and then used the background-subtracted median intensity of each spot for further statistical analysis. We initially analyzed two replicates of each clone within a chip separately and then discarded 5% of the weakest and saturated spots in each replicate. An intensity-dependent normalization method, namely, locally weighted linear regress (Lowess), was used to normalize the data for each replicate (Yang et al. 2002). Clones with ≤ 6 valid measurements were excluded from further statistical analysis, and two replicates per chip were averaged and used as the measurement of a clone for a given sample. Further, the normalized ratios of each dye-swap pair were averaged to get the final expression measurement of each gene. Significance Analysis of Microarray (SAM) (Tusher et al. 2001) was used to detect the significantly regulated genes, and a false discovery rate (FDR) value of 0.10 was used for all the brain regions.
