We used 54 pair-wise tissue samples, selected from 27 patients.
Genes that were differentially expressed in the pair-wise tissue comparisons were selected with the statistical approach described in [ 22].
The result can be expressed as a ratio between the two similarity measures, LTR+/LTR−, for every pair-wise tissue comparison.
Preprocessed images were visually inspected prior to case control comparison, with co-variation for age and global GM or WM (voxel-wise tissue classification relative threshold 0.8, cluster threshold k = 50, statistical threshold P < 0.05 with family-wise error correction).
These intervals were used to classify a given gene as differentially expressed if its replicate ratios in the pair-wise tissue comparisons were consistently (> 50%) outside the credibility interval thresholds.
On average, a similar fraction of concordant genes, i.e., genes that were also identified in the correspondent direct hybridization, was identified as differentially expressed by each indirect method (RefOligo: average of 58%% among the three pair-wise tissue comparisons; RefPool: average of 63 %; One-Color: average of 60%%).
The resulting 12 RA expression vectors were then used to compute pair-wise tissue sample correlations, of which the average values within and across the two tissues and technologies are plotted in figure 2. Both technologies show good reproducibility within the same tissue shown by the high correlation values.
We therefore defined the differentially expressed genes in the TaqMan data set as genes with an average fold change > 1.2 between pair-wise tissues, while for microarray data sets, the differently expressed genes were defined as p-value < 0.05 and average fold change > 1.2 between pair-wise tissues.
Differentially expressed genes were determined between pair-wise tissues for both microarray and TaqMan real-time PCR reference data sets as p-value < 0.05 based on a student's t-test.
For comparison of fold change between pair-wise tissues, scatter plots were generated between log2 Fold Change determined by microarrays and by TaqMan Gene Expression Assays (ΔΔCt), for every possible combination of tissue pairs, brain vs. liver, brain vs. lung and liver vs. lung,.
As a result of this study, we recommend using "reference" data sets generated by real-time PCR to evaluate critical aspects of microarray platforms, including signal detection threshold, fold change correlation between pair-wise tissues, profile correlation across multiple tissues, as well as sensitivity and specificity in signal detection and differential expression.
