Figure 1

Technical "block" effects due to cohort, chip-type, and laboratory (site). a. A color-coded correlation matrix among 201 AnCg samples. Red indicates high correlation between pairs of samples; blue indicates low correlation. Samples were ordered by technical batches, which in turn were defined by Cohort, Chip Type, and Site, as indicated below the heatmap. Throughout this paper the sample order in correlation heatmaps is the same from left to right as from bottom to top. Pearson's correlations were calculated by using all U133A transcripts. Rectangular "blocks" of high correlation, indicated in red, correspond to samples that are highly similar to each other in gene expression patterns. These block-block partitions coincide with the natural boundaries of experimental batches as indicated below the heatmap. D1 through D4 indicate Depression Cohorts 1 to 4, while S1 and S2 indicate Schizophrenia Cohort 1 and 2. b. The same correlation matrix after removing the block effect by median centering each block. The diagonal line of high values are self-self correlations, but the two off-diagonal lines of relatively high correlation values are between replicate chips for the same samples ran at two sites. The sample order and block partitions are the same as in 1a.