Preregistered cross-cohort evaluation of a metabolite-informed host-response score in adult sepsis
Abstract
This is a preregistered secondary analysis of public adult sepsis transcriptomic cohorts. We will evaluate a fixed, equally weighted five-gene host-response score whose gene directions were derived from infection-versus-reference comparisons without using mortality outcomes. The score is anchored to primary-source-verified gutMGene metabolite–host-gene edges but is not a direct metabolite measurement, microbiome-origin marker, causal score, treatment score, or clinical decision model. Registration A (OSF V9TFZ) had a null primary candidate lock (0/8 metabolites). Registration B (OSF 7YZEC) did not meet its primary KYNA intersection-union criterion (BH q=0.0717336). These null results will be disclosed and will not be reinterpreted as positive evidence for the present score. Gate 0 accessed public outcome labels only to verify sample/event counts, timing, covariate availability, and duplicate patients. No candidate score–outcome effect, odds ratio, P value, AUC, or cutpoint has been computed in the new cohorts before this registration.
