Fig. 4

Assessing phylogenetic origins of human divergence. a Preservation (Zsum) scores of three human co-expression modules in test expression datasets derived from human, non-human primate (NHP), and mouse. While human-mouse-module divergence is calculated by comparing the upper quartile (UQ) of the human Zsum scores against the mouse Zsum scores, this score can be further partitioned into human and primate specificity scores by considering the NHP UQ Zsum score. This subdivision allows us to determine these transcriptomic differences arose. b Human-mouse divergence, human specificity, and primate specificity scores for cell-type modules in human. Black box represents mean divergence/specificity score for each cell type. The cell-type color scheme applies to all panels of the figure—the microglial subclassification only applies from panel c onwards. c Human specificity and primate specificity scores for all cell-type modules in human, grouped by cell type. Black box represents mean specificity score for each cell type. d Summary plot showing modules with significantly increased human specificity. Module placement on the X-axis is based on its human-primate specificity ratio; modules to the left of the dashed line show greater primate specificity, whereas modules to the right show greater human specificity. Points are sized by their absolute divergence to mouse. Error bars represent the 95% CI from permuting across study in that species; * denotes regions with a CI that does not overlap zero. e Summary plot showing mean human-primate specificity ratio for each cell type calculated by averaging the module ratio scores for each cell-type class. Most cell types trend towards human specificity. Points are sized by their absolute divergence to mouse. Error bars represent the 95% CI from permuting across study in that species; * denotes regions with a CI that does not overlap zero. f Summary plot showing mean human-primate specificity ratio for each cell type, having derived cell-type markers from independent (“Methods and materials”; Additional file 12: Table S11) co-expression (square; Kelley at al., 2018), single-cell RNA sequencing (triangle up; Lake et al., 2018), and cell-sorting-based experiments (triangle down; Zhang et al., 2016), which are in general highly concordant. Points are sized by their absolute divergence to mouse. Trends observed for GTEx-derived modules (e) are mirrored in these additional datasets