Mechanism-Driven Astragalus Pathways Signatures for Risk Stratification in ICU Sepsis: A Translational Bioinformatics Approach

Background: Astragalus, a traditional Chinese medicine, exerts pleiotropic pharmacology through multi-component, multi-target mechanisms. We curated a mechanism-anchored prior—“Astragalus → mechanistic nodes → representative genes”—from the literature and evaluated its pre-specified primary association in an ICU sepsis cohort.
Methods: We systematically searched PubMed, Embase, and Web of Science, identifying 7,774 records; 2,222 studies met inclusion after tiered screening. Evidence was stratified by experimental validation and effect direction to build two gene-set resources: an integrated pathway signature and node-specific subsets. In GSE65682 (ICU whole-blood transcriptomes; n=479), we computed single-sample scores using mean-Z transformation and singscore. Age-adjusted logistic regression estimated associations with 28-day mortality, controlling multiplicity with Benjamini–Hochberg false discovery rate (FDR). Restricted cubic splines (RCS) probed nonlinearity for the A-NLRP3 node.
Results: The integrated signature showed a directionally consistent protective trend across scoring methods; mean-Z estimates often reached or approached multiplicity-adjusted significance, while singscore yielded concordant but attenuated effects. At the node level, NLRP3 retained a robust protective association after FDR correction (OR=0.70, 95% CI 0.57–0.86; P=6.02×10⁻⁴; q=0.0037) and displayed a linear dose–response (Poverall=0.017, Pnonlinear=0.986). NF-κB and TLR4 exhibited nominally protective associations that did not survive FDR correction; several nodes showed method-dependent signals.
Conclusions: Mechanism-driven Astragalus pathway signatures associate directionally with protection in ICU sepsis, with NLRP3 inflammasome regulation emerging as a stable mechanistic anchor. This reproducible framework translates a multi-target botanical into clinically interpretable biomarkers and supports precision applications in critical care.
Keywrods: Sepsis; Intensive Care Units; Astragalus; Inflammasomes; NF-kappa B; Toll-Like Receptor 4; Mitochondria; Transcriptome