Fig. 3
From: Modeling and cleaning RNA-seq data significantly improve detection of differentially expressed genes
Comparison of different data cleaning procedures. A Histograms of p-values of DEGs after application of different filters. Histograms in the background (grey) represent the p-values of raw data, in foreground of filtered data (more filters are in Additional file 1: Fig. S1). B Histogram of ratios of p-values before and after cleaning with RNAdeNoise. Asymmetry against negative values shows higher significance of DEGs after cleaning. C DEGs identified by all filters, by two or more filters and DEGs unique to each filter. RNAdeNoise detects most of genes detected by other methods plus many new genes. Presented are genes with moderate expression on which filtering has a strongest effect (Additional file 1: Table S2). D Average per-gene dispersion (EdgeR parameter tagwise.dispersion) and LogCPM for raw data and data cleaned by different filters. E Number of detected DEGs as a function of the filtering strength