Skip to main content
Fig. 3 | Genome Biology

Fig. 3

From: Do count-based differential expression methods perform poorly when genes are expressed in only one condition?

Fig. 3

Performance for zero-in-one-condition simulation. a ROC curves and b true positive rate (TPR) versus ‘achieved’ FDR curves of DE methods for the simulation dataset with zero-counts introduced as the true DE genes (overall performance of three simulations). The achieved FDR is the actual rate of false discoveries at the corresponding cutoff and this rate should ideally be controlled at the desired level. For the ROC curves, the cross on each curve represents the method’s TPR at the (estimated) 5 % FDR cutoff. For the TPR versus achieved FDR curves, points are plotted at the following cutoffs: 0.01, 0.05 and 0.1. Filled-in points represent that the method has correctly controlled the error rate at the cutoff. FDR false discovery rate, FPR false positive rate, roc receiver operating characteristic, TPR true positive rate

Back to article page