Population scale proteomics enables adaptive digital twin modelling in sepsis

Sepsis is one of the leading causes of mortality in the world. Currently, the heterogeneity of sepsis makes it challenging to stratify patients on admission to the emergency department, hindering the implementation of precision medicine. Here, we acquired clinical parameters and blood plasma proteome profiles for 3009 patient samples to develop a framework for digital twin-based modelling in sepsis. Already on admission to the emergency department, the model can accurately diagnose sepsis and infection, while providing prognostic predictions for persistent organ dysfunction, mortality, and intensive care unit admission. Furthermore, the model can suggest therapeutically actionable pathways based on patient-unique proteome trajectories, determine the infection loci and pathogen type, and identify which patients may benefit from vasopressor treatment at the time-of-admission to the emergency department. The framework has the potential to advance precision medicine in sepsis and provides a generalizable approach for data-driven disease modeling and clinical decision support.