Ancestry-Associated Pathway Enrichment and Network Architecture in Severe Asthma: Airway Remodelling, Epithelial Barrier and Sex-Adjusted Analysis

Abstract Background Systems-level transcriptomic analyses may identify coordinated biological programmes that are not apparent from single-gene testing, but pathway and co-expression results from whole blood are sensitive to sex, cell composition and sample size. Objective To characterise ancestry-associated pathway enrichment, sample-level pathway scores and curated-gene co-expression structure in severe asthma. Methods We analysed RMA-normalised whole-blood microarray profiles from 24 participants with severe asthma in GSE69683: 12 recorded as black_african and 12 as white_caucasian, with eight females and four males in each group. Gene-level expression was derived from 21,653 unambiguously mapped genes. Pre-ranked GSEA used the sex-adjusted ancestry t statistic across ten curated asthma pathways; ssGSEA scores were compared using sex-adjusted ordinary least squares. Spearman networks among 128 measurable curated genes used a pre-specified |rho| >= 0.60 and FDR < 0.05 threshold. Results Pre-ranked GSEA identified Epithelial Barrier Function (NES − 1.871, FDR 0.0152) and Airway Remodelling (NES − 1.766, FDR 0.0193), both enriched towards the European-ancestry group. Adjusted ssGSEA identified Airway Remodelling (FDR 0.0001), Bronchial Hyperresponsiveness, Innate Immunity/TLR Signalling and Steroid Response (each FDR 0.0348), again with higher mean scores in the European-ancestry group. The full network retained 173 of 8,128 tested pairs and connected 81 genes. COL1A1 and ARRB2 had the highest degree (13 each), while SPDEF had the highest betweenness; subgroup networks were too sparse for reliable comparative network structure claims. Conclusions Airway Remodelling was the only pathway supported by both enrichment approaches, while epithelial-barrier, bronchial hyperresponsiveness, innate/TLR and steroid-response findings depended on the analytical question being asked. The co-expression results identify descriptive network-central candidates, but sparse subgroup graphs preclude reliable comparative network inference. Whole-blood cell composition, unavailable treatment outcomes and the small selected cohort constrain interpretation; independent replication in larger ancestry-diverse cohorts with cell-composition and richer clinical data is required.