Extended Data Fig. 6: Application of GEP usage to lineage branching and HSC lineage dominance.

(a) Based on the pair-wise comparisons between the six lineage GEPs displayed in Fig. 2a, volcano plots of the negative-log of Bonferroni corrected p-value of the Pearson Correlation coefficient versus the Correlation coefficient itself are displayed for HSCs, MPPs, and Progenitors in all age groups. (b) For fetal, adolescent, and adult MPPs and progenitors identified in²⁶, two-sided Pearson Correlation coefficients and Bonferroni corrected p-values of the coefficient were calculated for all pair-wise comparisons between the six lineage GEPs. These adjusted p-values were then compared to the adjusted p-values of the lineage anti-correlations found to be significant in this current study. (c) A schema is presented describing two methods (division- and subtraction-based) for calculating unilineage dominance score using either lineage GEP usage from single cell transcriptome data or lineage output in clonal outgrowth assays in culture. (d,e) Using a (d) division-based or (e) subtraction-based methodology, unilineage dominance scores were calculated in individual fetal, adult, and elderly HSCs using either lineage GEP expression or lineage output in a clonal outgrowth assay supporting lymphoid, monocyte/dendritic cell, granulocyte, erythroid, and megakaryocyte differentiation. The median unilineage dominance scores (indicated by horizontal lines on the violin plots) from GEP usage and clonal outgrowth for each age were then compared and lineage regressions were performed on the relationship between scores calculated with GEP usage or clonal outgrowth data. Displayed coefficients of determination indicate that for HSCs, unilineage dominance scores calculated from GEP usage and clonal outgrowth data more closely reflect each other when using a division-based methodology. (f) Using both the division-based and subtraction-based methods for calculating unilineage dominance, scores for individual HSCs and individual progenitors were calculated, with median indicated by the horizontal line on each violin plot. Significance testing for the difference between HSCs and progenitors in both the division-based and subtraction-based plots yielded p-value<2.2x10-¹⁶ by two-sided Wilcoxon Rank-Sum testing. The increased unilineage dominance score in progenitors compared to HSCs indicates the subtraction-based methodology is preferred for determining magnitude of lineage commitment.