Figure 2.

285 Sex-dependent DEGs confirmed as important features for cell sex classification model a, Schematic of feature selection using the VTA training data partition. b, Heatmap of log2 fold changes of genes significantly differentially expressed (adjusted p < 0.05) in at least one cell type (408 genes), color scale is capped from −2.5 to 2.5. Genes are ordered along the x-axis by ranked mean log2 fold change across cell types. c, Boxplots of gene importance scores as determined through 2000 iterations of the Boruta algorithm. Genes are ordered along the x-axis by median importance score. Boxplots are colored by the final importance decision by Boruta. (”Shadow” features are generated by Boruta from randomly shuffled gene data, used to evaluate the importance of real feature data). d, Violin plots of gene expression levels, within cell type split by sex, of the two highest confirmed importance male (ENSRNOG00000060617 (Uty), Ddx3) and female (ENSRNOG00000065796 (Xist), Kdm6a) biased, and the two lowest importance rejected (Fam210a, Trabd2b) sex-dependent DEGs determined by Boruta.