Deriving consensus sepsis clusters via goal-directed subgroup identification in multi-omics study.

Area of Science:

• Critical Care Medicine
• Genomics and Systems Biology
• Translational Bioinformatics

Background:

• Sepsis exhibits significant heterogeneity, complicating treatment and leading to poor therapeutic outcomes.
• Current patient subtyping methods using single-omics or unsupervised clustering lack reproducibility and clinical relevance.
• Developing effective sepsis therapies is hindered by imprecise patient classification.

Purpose of the Study:

• To introduce a novel Goal-Directed Subgroup Identification (GD-SI) framework for sepsis patient stratification.
• To optimize patient classification for differential treatment responses using longitudinal multi-omics data.
• To enable precision trial design and personalized sepsis management.

Main Methods:

• Integrated longitudinal multi-omics data (transcriptomic, proteomic, metabolomic, phenomic) from 1327 subjects across 43 hospitals.
• Employed a supervised multi-omics integration approach anchored to treatment-effect optimization.
• Validated the framework's generalizability in international critical care databases (MIMIC-IV, ZiGongDB).

Main Results:

• Achieved substantial cross-omic concordance in patient subgroup discovery.
• GD-SI successfully predicted differential treatment responses for fluid resuscitation and ulinastatin therapy.
• Stratified patients based on GD-SI benefit scores showed significant survival differences.
• External validation confirmed the prognostic generalizability of the framework.

Conclusions:

• The GD-SI framework reconciles sepsis biological heterogeneity with clinical actionability.
• This approach offers a scalable infrastructure for precision trial design and personalized sepsis management.
• Omics-driven, goal-directed stratification holds translational potential to advance critical care therapeutics.