Skip to main content


Fig. 4 | BMC Genomics

Fig. 4

From: aFold – using polynomial uncertainty modelling for differential gene expression estimation from RNA sequencing data

Fig. 4

RNA-Seq analysis of the qRT-PCR validated data sets. Analysis is performed on three data sets: ABRF, SEQC and MAQC-II. PrimePCR is used to define true and false positives (TPs, FPs), TPR and FPR are defined as in Fig. 1. (a) ROC analysis. Solid lines show the results for the RNA-Seq methods with their integrated normalization procedures. Dashed lines (except diagonal) show the results under qtotal for all methods except for the two methods, aFold and ABSSeq, which use qtotal as default and are thus shown as solid lines. Qtotal improves the performance of most methods. (b) Sensitivity analysis. Sensitivity is calculated as the ratio between the number of true DE genes under adjusted p value < 0.05 and the total number of true DE genes, inferred from PrimePCR. The empirical false discovery rate (eFDR) is calculated as FPs/(TPs + FPs) under adjusted p value < 0.05. qtotal improves either eFDR or sensitivity or both when applied with the tested DE methods. Filled and open circles represent methods with default normalization approach and qtotal, respectively

Back to article page