Supplemental Figure 3. Spearman correlation produces similar results as the Pearson correlation for measuring gene-gene co-expression in the four organisms. We tested the possibility that the Pearson correlation was sensitive to extreme values in the microarray datasets by also computing Spearman correlations for every pair of genes in each of the four organisms. We found that the Spearman rank correlation, which is robust against the presence of such extreme values but which is also a less sensitive measure of correlation than the Pearson correlation, agrees well with the Pearson correlations. The agreement is reflected in the strong positive linear relationship between the two measures of correlation. A-D. Pearson versus Spearman scatter plots for each of the four organisms. Each dot in each of the plots shows the Spearman (x-axis) and the Pearson (y-axis) correlations computed for a single gene-gene pair.