Fig. 2
From: Large-scale labeling and assessment of sex bias in publicly available expression data
Breakdown of study sex labels. Alluvial diagram showing the breakdown of study sex labels in the metadata (left of each panel) and after expression based labeling (right of each panel). The flow is colored by the initial metadata labels and helps trace whether there is a “change” in labels. For the majority of studies with metadata labels, the labels match the inferred expression labels. The results are shown for both human and mouse (columns) and in microarray and RNA-seq (rows). Gray indicates that a study is missing sex labels (for studies with up to 60 samples, more than half of the labels are missing: for studies with greater than 60 samples, there are fewer than 30 labels), dark blue means the samples in the study are female-only, dark orange is male-only, and pink is mixed-sex. Mostly male (light orange) and mostly female (light blue) indicate that more than 80% of the samples labeled in that study are of that sex. For the metadata breakdown, the numbers of studies in mostly-male and mostly-female categories was small (e.g. 71 mostly-female and 87 mostly-male studies in human microarray) and were grouped into mixed sex for ease of visualization. (For a similar figure with sample sex breakdown see Additional file 1: Fig. S1.)