Figure 2

Comparison of different methods. The area under the receiver operating characteristic curve is used as the performance measure. The box plots summarise the results obtained across three independent synthetic datasets for four different simulation settings. Each dataset included 10K genes and results across 2, 4 and 8 biological replicates are reported. DGEclust shows improved performance in comparison to other methods in all of the examined cases, particularly in the presence of a large proportion of differentially expressed genes (30%) and small sample sizes (n=2). The inclusion of non-over-dispersed genes significantly improves the performance of all methods, but the inclusion of outliers has the opposite effect. Still, DGEclust remains top-ranked among the alternative methods with respect to AUC scores. AUC, area under the receiver operating characteristic curve; DE, differentially expressed; ROC, receiver operating characteristic.