SepsisShearome: analysis code for a meta-analytic endothelial shear-stress atlas of sepsis endotypes (v6 submission)

S21) — EC protective module −0.57 (P=0.028, BH-P=0.085), composite ESS +0.65 (P=0.028), composite elevation confined to the endothelial compartment; Figure 6 and D21 tables regenerated; Bonferroni-adjusted null-matched P qualified by null resolution. 2026-09-22 fourth-round upgrade (230 scripts): pre-specified interaction question resolved in the largest atlas (ESS×C4 on 28-day mortality, n=1,689, unadjusted: OR 1.07, 0.82–1.39, P=0.60; four-subtype LRT P=0.70) with three-cohort random-effects pooling (OR 1.08, 0.15–7.64, I²=0%) — effect modification resolved as absent; C4×ESS quadrant mortality gradient reproduces the two-axis logic (dual-high 34.2% versus both-low 18.3%, trend P=9.0×10⁻⁶; within-C4 quartile dose–response 21.4%→37.1%); outputs in D32_interaction_atlas (p72).2026-09-25 fifth-round review response (248 scripts): p84-p88 added - GSE171668 unit/paired analyses with seven degenerate cross-compartment backfill units excluded and re-run; overlap-excluded three-cohort interaction pooling (OR 1.14, 0.65-1.99); compartment x disease interaction model (composite ESS endothelial effect +0.73, uncorrected); GSE310929 constituent-series stratification (within-series z: KW H=100.7; C4 median above other subtypes in 9/9 multi-subtype non-overlap series; series-adjusted gradient P=1.7e-7); adjacent-stress-programme null battery extended to both validation cohorts - the separation from the hypoxia-directional axis is discovery-specific (P=0.278/0.250) and reported as such; composition-drift simulation re-estimated at 10,000 draws x 3 seeds (seed-median P99.9 0.56-0.57). requirements_lock.txt and R_lock.txt dependency locks included; all T_p80-T_p88 result tables archived with the package.