Identification of outlier samples by correlation analysis and scalar factors. (a) Plotting the average correlation for each sample from pairwise comparisons to all other samples (y-axis). The samples were sorted according to their average correlation (x-axis). We used an average correlation of 0.34 as a cutoff (marked with a dashed line). (b) Comparison of the average correlation (x-axis) with the scalar factor used in the global scaling procedure (y-axis). Many of the samples with low average correlations had been rescaled using high scaling factors, indicating that they might have had poor hybridizations. Again, the dashed line displays the average correlation cutoff.