An NLRP3 inflammasome–anchored Astragalus mechanistic prior yields a mortality-associated transcriptomic signal in ICU sepsis: a secondary analysis with exploratory cross-cohort assessment

OBJECTIVE AND DESIGN: We developed ASTRA, a literature-informed, author-defined Astragalus mechanistic prior, and examined whether integrated and node-level whole-blood transcriptomic scores were associated with 28-day mortality in ICU sepsis. METHODS: The primary analysis included 479 adults with sepsis from GSE65682; an exploratory cross-cohort directional assessment included 51 Day-1 patients with septic shock from GSE95233. Mean-Z scores were evaluated using age-adjusted logistic regression with false-discovery-rate correction. Sensitivity analyses included singscore, restricted cubic splines, label permutation, leave-one-gene-out analysis, correlations with IL1B and IL6 mRNA, and adjustment for transcriptome-derived myeloid-cell composition. RESULTS: The integrated ASTRA score showed a nominal inverse association with mortality but did not survive false-discovery-rate correction. The NLRP3-related three-gene score showed the strongest association in GSE65682 (OR per 1-SD increase, 0.70; 95% CI 0.57-0.86; q = 0.0072) and remained directionally stable in leave-one-gene-out analyses. It correlated positively with IL1B and IL6 mRNA and with MCP-counter monocytic-lineage and neutrophil scores. Joint MCP-counter adjustment attenuated the association to an OR of 0.79 (95% CI 0.58-1.06), whereas xCell neutrophil adjustment strengthened it to an OR of 0.64 (95% CI 0.50-0.83), indicating method-sensitive dependence on estimated cell composition. In GSE95233, the age-adjusted NLRP3 estimate was directionally concordant but imprecise (OR 0.68; 95% CI 0.37-1.23). CONCLUSIONS: ASTRA provides a reproducible framework for mechanism-guided transcriptomic scoring in sepsis. The NLRP3-linked signal represents an observational whole-blood transcriptomic state that partly tracks estimated myeloid-cell composition and concurrent inflammatory transcription. It does not establish protective inflammasome activity, Astragalus efficacy, or pharmacological target engagement.