A Machine learning model for predicting sepsis based on an optimized assay for microbial cell-free DNA sequencing.

Area of Science:

• Molecular diagnostics
• Machine learning
• Sepsis research

Background:

• Sepsis diagnosis remains challenging, requiring rapid and accurate methods.
• Current diagnostic techniques may have limitations in early detection and specificity.

Purpose of the Study:

• To develop and validate a machine learning model for sepsis diagnosis.
• To integrate an enhanced molecular diagnostic technique for improved accuracy.

Main Methods:

• Prospective enrollment of patients suspected of sepsis.
• Development of a machine learning model using feature selection and cross-validation.
• Utilized optimized mNGS assay and microbe-specific cell-free DNA (CPM) as a detection signal.
• Employed SHAP method for feature interpretability.

Main Results:

• A random forest classifier achieved high performance (AUC 0.918, F1 0.856) in the training set.
• The model demonstrated good prediction performance in the testing set (AUC 0.85, average precision 0.91).
• Key features identified include PCT, CPM, CRP, ALB, SBPmin, RRmax, CREA, PLT, and HRmax.

Conclusions:

• Machine learning approaches combined with optimized mcfDNA sequencing can accurately predict bacterial sepsis.
• The developed model shows potential for clinical application in sepsis diagnosis.