Figure 3

False discovery curves for all methods across the first 1,000 discoveries. The illustrated false discovery curves are averages over three independent repetitions of each synthetic dataset. DGEclust clearly keeps a lower number of false discoveries in comparison to the other methods in all cases. There is a single exception in the presence of outliers and at large sample sizes (n=8), where DESeq2 appears to be marginally better than DGEclust over the first 500 discoveries. DE, differentially expressed.