Figure 2.

ROC curve. This graph displays the number of false negatives (differentially expressed genes (DEGs) not included) versus the number of false positives (non-DEGs included) found on top lists of increasing sizes, expressed as proportions of the total number of genes. The distance C gives an optimal value of equation (2). A method whose ROC curve lies below that of another method is preferable, as it will give more DEGs and fewer non-DEGs on any top list of any size, as explained in Additional data. Hence method 1 is preferable to method 2.