Fig. 4

R² of measured versus expected log2 fold-changes between samples C and D. a Gene expression level influenced accuracies of fold-change estimation. R² values were computed from the expected and measured log2 fold-change values between samples C and D for each pipeline using different gene sets grouped by average gene expression levels. The first gene set, labeled “Total RNA”, includes all genes. The subsequent gene sets include only the genes with the top 1%, top 10%, top 25%, or bottom 75% expression levels, as indicated. Bars are color-coded by pipelines. b Gene lengths influenced accuracies of fold-change estimation. Genes from each pipeline were grouped by their gene lengths into four quantile groups. For each quantile group, R² values were computed from the expected and measured log2 fold-change values between samples C and D. Bars are color-coded by pipelines. Coefficients of determination (R²) were computed by R2 function from R caret package [35]. Negative R² values indicate exceptionally bad fold-change predictions [23] from the software as illustrated in Additional files 7 and 8, where the fold-change prediction do not fit well to the samples mix-ratio