Fig. 3

Comparison of decision boundaries in the real data-based simulation study. The log2 fold changes of gene expression levels are shown for genes in the simulation in Fig. 2. The genes that are ranked in the top 20 % by each method are deemed as DEGs. Correctly detected DE genes, false positive, false negative and true negatives from each method are shown in green, black, red and grey, respectively. The difference in decision boundaries shows that genes with moderate but consistent signals are more likely to be identified by our method