In this first step, each angular view is corrected for the dark current of the detector and the background is normalized using the average of the acquired white-field images.
Median foreground signal intensities (no background subtraction) were normalized using Locally Weighted Linear Regression (LOWESS) within the R statistics package (version 2.7.2) using the "VSN" method in limma (version 2.14.7) to remove signal intensity-dependent dye bias.
After correcting background noise by subtracting median background signals, data were normalized using Loess method in Partek Genome Suite version 6.5.
In brief, background-corrected data were normalized using quantile normalization and summarized by median polish.
After background correction, data were normalized using the ABI 1700 Chemiluminescent Analyzer first by feature, then by spatial effects in the slide.
The background-corrected data were normalized using the Gene Expression Pattern Analysis Suite v3.1 (GEPAS, http://www.gepas.org) and the global loess method [ 69].
After subtracting background noise, the intensity data were normalized using a 75th-percentile normalization in genespring software version 11.0.2 (Agilent Technologies, Santa Clara, CA, USA).
Background-corrected hybridization signals were normalized using R/MAANOVA version 0.98.8 implemented in the R program (www.r-project.org).org
The background-corrected intensity data were normalized using the intensity-dependent LOWESS method to remove the dye bias within each array as recommended by the manufacturer (Agilent, USA).
Background-subtracted raw data were normalized using the MIDAS pipeline (TM4, TIGR Genomics, Rockville, MD) according to Sioson et al. (2006) with the following steps: total intensity normalization, LocFit (LOWESS), standard deviation regularization and low intensity trim [ 25].
