Fig. 1
From: Large-scale labeling and assessment of sex bias in publicly available expression data
Schematic of study analysis. We constructed our dataset from all human and mouse microarray and RNA-seq data in refine-bio. We then extracted metadata labels for sex, cell line, and drug, and built elastic net models to infer sample sex from gene expression. Using the metadata and inferred sex labels, we investigated the sex breakdown of samples by organism, as well as cell line “sex” complexity (e.g., Y chromosome loss). We also looked for the presence of mislabeled data by examining concordance between predicted and actual sex labels, and at sex bias in drug studies (e.g., underrepresentation of females in studies of nervous system drugs)