Figure 5

False discovery rates for all methods at a pre-specified significance threshold. The box plots summarise the FDRs obtained across three independently obtained simulated datasets at four different simulation settings at an imposed significance level of 10%. The ability of all methods to control their FDR increases with the sample size. In all cases, DGEclust demonstrates excellent control over its FD, particularly at small sample sizes (n=2). Interestingly, in the presence of outliers, DGEclust is the only method that keeps its FDR at or below the pre-specified threshold at all sample sizes. DE, differentially expressed; FD, false discovery; FDR, false discovery rate.