Integrative analysis of mortality risk in SFTS using machine learning and genetic approaches

Area of Science:

• Infectious Diseases
• Computational Biology
• Genetics

Background:

• Severe fever with thrombocytopenia syndrome (SFTS) is a significant tick-borne viral disease in East Asia with high mortality rates.
• Early identification of high-risk patients is crucial for timely intervention and improved outcomes.

Purpose of the Study:

• To develop and validate a machine learning (ML) model for predicting mortality risk in SFTS patients.
• To identify key clinical and genetic factors associated with SFTS mortality.
• To explore potential causal pathways influencing SFTS severity.

Main Methods:

• Utilized clinical data from 107 confirmed SFTS cases.
• Developed and compared various ML models, including extreme gradient boosting (XGBoost).
• Employed Boruta algorithm for feature selection and SHAP plots for variable importance analysis.
• Conducted Mendelian randomization to investigate causal relationships.

Main Results:

• The XGBoost model demonstrated high predictive performance with an AUC of 0.950.
• Fifteen robust features, including IL-6, TNF-α, PLR, and AST, were identified as key predictors.
• SHAP analysis provided insights into the contribution of each feature to mortality risk.
• Mendelian randomization suggested associations within specific metabolic-inflammatory pathways, though definitive causal links were not established for all factors.

Conclusions:

• The developed ML model offers a powerful tool for early and accurate prediction of SFTS mortality.
• Identifying key predictive variables enhances understanding of SFTS pathogenesis.
• Further research into identified metabolic-inflammatory pathways may reveal novel therapeutic targets.