Fig. 6

Overall performance of workflows and pairwise-comparison using refine.bio datasets. The boxplots show the aggregate accuracy of all coexpression networks resulting from each individual workflow using SRA datasets in refine.bio, evaluated based on the tissue-naive gold standard. The performance of each workflow is presented as boxplots (without outliers) that summarizes the log2(auPRC/prior) of each workflow, where auPRC is the area under the precision recall curve (see the “Methods”). The workflows are ordered by their median log2(auPRC/prior). The heatmap shows the relative performance of pairs of workflows (rows and columns) compared to each other for the refine.bio SRA datasets based on the tissue-naive gold standard. The color in each cell (row, column) represents the proportion of datasets for which the workflow along the row has a higher log2(auPRC/prior) than the workflow along the column. Comparisons that are statistically significant (corrected p < 0.01) based on a paired Wilcoxon test are marked with an asterisk. Figure S8 contains these plots based on the tissue-aware gold standard