Fig. 6

Checking common errors of RNA-seq data analysis using side-by-side boxplots and MDS plots. Side-by-side boxplots and MDS plots are popular plotting tools for RNA-seq analysis. This figure shows these traditional visualization methods applied to the soybean cotyledon data before sample switching (left half) and after sample switching (right half) [27]. We cannot suspect from the right boxplot that samples S1.3 and S2.1 have been swapped (subplots A). This is because all six samples have similar five number summaries. For the MDS plots, we do see a cleaner separation of the two treatment groups across the first dimension in the left plot than in the right plot (subplots B). However, taking into account the second dimension, both MDS plots contain three clusters, with sample S1.1 appearing in its own cluster. Without seeing one distinct cluster for each of the two treatment groups, it is difficult to suspect that samples S1.3 and S2.1 have been swapped in the right MDS plot (subplots B). We can only derive clear suspicion that the samples may have been switched by using less-popular plots that provide gene-level resolution like with the scatterplot matrix from Fig. 5