Total RNA was treated with DNase I (RNeasy - Qiagen) before labeling
Data processing
We employed in our analyses the Median-based Trimmed Mean density value (MTM) for each spot by Arrayvision. The reported value represents the mean of all the pixels remaining in a target, after first removing pixels with density values that exceed four median absolute deviations (MADs) above or below the median intensity of the target. This measure moderates the influence of image artifacts (e.g., dust particles) on density estimation. The MTM density for each spot was subtracted by the individual background surrounding the spot (size 4 pixels). Spot quantization matrixes (the �raw� output of the image analysis software) were obtained from all hybridizations. Measurements correspond to the raw data generated by the Image analysis software. We have consolidated in a single tab-delimited text file all measurements obtained in all hybridizations in order to facilitate subsequent normalization and averaging of signal intensities from each tissue. Microarray data intensities from spots representing controls were removed and raw quantization matrixes from each tissue samples were normalized using a robust locally weighted linear regression algorithm (lowess), implemented into the R package, to correct for systematic biases on the data originated from small differences in the labeling and/or detection efficiencies between the fluorescent dyes. Next, in order to make experiments comparable, expression profiles were scaled to a reference, based on the trimmed mean intensity of all intensities measured in each sample that were expressed in at least one tissue, where the highest and the lowest 20% of the values were excluded from the calculated mean. A parallel analysis was performed to detect spots with signal intensity above the detection cutoff, which was determined by the average intensity plus 2 standard-deviations of 1,198 negative controls present in microarray platform. The raw intensity of each spot was compared to the detection cutoff of the corresponding dye. A spot was considered �expressed� if all 4 signals from a tissue were above the detection cutoff. For statistical analyses, Significance Analysis of Microarray (SAM) approach was employed using as parameters: multi-class response, 1000 permutations, K-Nearest Neighbors Imputer, and false discovery rate (FDR) 0.00001%; analysis of variance (ANOVA) implemented in the SpotFire Decision Site for Functional Genomics (SpotFire Inc.) with a cutoff of p less than 0.001. Gene sets identified by both SAM and ANOVA were combined in order to identify a more restrict set of spots that showed statistically significant expression changes in a tissue by both analyses.
Raw data - Background-subtracted Median-based Trimmed Mean density value (MTM) extracted from Arrayvision software (Version 8.0, Imaging Research Inc.)