Figure 2 : The selection of genes for the final between-condition normalization in MedianCD and SVCD normalization was crucial to preserve the variation between conditions.

From: Variation-preserving normalization unveils blind spots in gene expression profiling

Figure 2

Panels show the detected variation as a function of the number of genes used in the between-condition normalization, for the real dataset (a), synthetic dataset with differential gene expression (b), and synthetic dataset without differential gene expression (c). Between-condition variation is represented as the standard deviation of the within-condition mean averages (averages of sample means, for all samples of the condition). See Supplementary Fig. S1 for results using within-condition median averages, with similar behavior. Each point in each panel indicates the variation obtained with one complete normalization (black circles, MedianCD normalization; blue circles, SVCD normalization). Genes were selected in two ways: randomly (empty circles) or in decreasing order of p-values from a test for detecting no-variation genes (filled circles). Big circles show the working points corresponding to the results depicted in Fig. 1j–o, which were chosen automatically. Black dashed lines show references for n−1/2 decays, with the same values in all panels.