Fig. 1: Study design and multi-omics analysis.
From: Deriving consensus sepsis clusters via goal-directed subgroup identification in multi-omics study
The Chinese Multi-omics Advances In Sepsis (CMAISE) cohort prospectively enrolled 1327 participants (including 47 healthy controls) from 43 tertiary hospitals across mainland China. Longitudinal blood sampling (days 1, 3, and 5) yielded 3895 patient-days of phenomic data, with matched transcriptomic (n = 2,776), proteomic (n = 468), and metabolomic (n = 457) profiles. Key analytical steps included: (1) Unsupervised Consensus Clustering: Applied to each omics layer (transcriptome, proteome, metabolome, phenome) to identify sepsis subtypes, followed by cross-omics cluster concordance analysis. (2) Goal-Directed Subgroup Identification (GD-SI): Implemented via LASSO-penalized regression to optimize subgroups for differential treatment effects of ulinastatin (UTI) and fluid resuscitation strategies. (3) Subgroup Validation: Cross-omics concordance of GD-SI-derived clusters and external validation of fluid response subgroups in public critical care datasets (e.g., MIMIC-IV, eICU-CRD). CMAISE Chinese Multi-omics Advances In Sepsis, UTI ulinastatin, GD-SI goal directed subgroup identification, LASSO least absolute shrinkage and selection operator, HR hazard ratio, MIMIC medical information mart for intensive care, eICU-CRD eICU collaborative research database.
