Fig. 4
From: DegNorm: normalization of generalized transcript degradation improves accuracy in RNA-seq analysis
Differential expression (DE) analysis in data sets that had no true differential expression. Results are shown for SEQC-AA, PBMC-S01, GBM R10 + R6, GBM R10 + R4, DLPFC Br1729 Ribo-Zero, and breast tumor data. a–f Coefficient of variation (CV) vs. mean read counts (in log scale): compared are results from proposed DegNorm pipeline and other methods including upper quartile (UQ), RUVr, and TIN. The mean counts for each data were scaled to 0–1 range by a linear transformation (i.e., (Xi − minj(Xj))/maxj(Xj) where Xi is the log of the mean count for gene i). The CV curve was generated using the R built-in function smooth.spline. The beanplot under the CV plot shows the density of log of mean read counts from DegNorm (the densities of read counts from other normalization methods are similar and not shown). g–l Empirical cumulative distribution function (ECDF) of the p value from DE analysis. The RUVr results were generated using the RUV-seq package. For TIN method, we followed Wang et al. [13] with details described in the uploaded R Markdown fileÂ