Fig. 4: Association of program regulation with blood traits.
From: Causal modelling of gene effects from regulators to programs to traits
a,b, Overview of our pipeline for the analysis to find the trait-relevant programs. c–e, Program burden effects (x axis) and regulator–burden correlation (y axis) of 60 programs in three blood traits: MCH (c), RDW (d) and IRF (e). Programs with significant associations after Bonferroni correction (P < 0.05/60) are coloured. Pathway annotations of representative programs are labelled. For annotations of other programs, see Supplementary Table 3. The P values for program burden effects are from the permutation test and are two-sided. The P values for regulator–burden correlations are from the linear regression and are two-sided. f, Schematic for the concordant and discordant patterns between program burden effects and regulator–burden correlation. P, program; R, regulator; T, trait. g, Co-regulation patterns between programs. Each dot represents a gene that has significant regulatory effects on the G2/M phase program. The gene effect size on the program activity was calculated by comparing the program usage of cells between perturbed cells and control cells using linear regression (βx→P; Methods). The lines and their 95% confidence intervals are from locally estimated scatterplot smoothing. h, A summary of signs of regulatory effects on the programs. i, Average γ and its standard errors for 115 genes in RA and 154 genes in RB MCH. j, Program association with MCH in GWAS–trans-eQTL analysis (Methods) and LoF–Perturb-seq analysis.
