Fig. 4

Impact of individual methods on performance of workflows. Each bar in the two barplots, corresponding to a specific method, shows the proportion of times (x-axis) that workflows including that particular method (y-axis) were significantly better than other workflows. The barplots correspond to performance for the a GTEx and b SRA datasets evaluated on the tissue-naive gold standard. In order to make the comparison of between-sample normalization methods fair, workflows also including CPM, RPKM, or TPM were left out because it is not possible to pair them with CTF/TMM/CUF/UQ normalization. Similarly, CTF/TMM/CUF/UQ methods are not included for “no within-sample normalization” (NO–WI). Figure S6 contains these barplots based on the tissue-aware gold standard