HBV-Specific T Cell Responses for Short-Term Hepatitis Progression Prediction: A Comparative, Interpretable Machine...

Area of Science:

• Immunology
• Hepatology
• Machine Learning

Background:

• Hepatitis B virus (HBV)-specific T-cell responses are crucial for viral control and immunopathology.
• Existing prediction models for hepatitis progression often lack biological interpretability due to limited inclusion of T-cell data.
• There is a need for advanced models that integrate diverse data sources for improved prediction accuracy.

Purpose of the Study:

• To develop and evaluate machine learning (ML) models for predicting hepatitis progression in HBV-infected individuals.
• To assess the impact of integrating HBV-specific T-cell responses with clinical data on predictive performance.
• To create an interpretable model for early risk stratification and proactive monitoring.

Main Methods:

• A cohort of 479 patients was analyzed, with data split into training and testing sets.
• Machine learning models were benchmarked using clinical data and HBV-specific T-cell immune responses.
• An XGBoost model was selected for its robust performance in cross-validation and independent testing.

Main Results:

• Integrating clinical indicator features (CIF) and HBV-specific T-cell features (STCF) significantly enhanced predictive performance.
• An XGBoost model utilizing six features (three CIF, three STCF) achieved high AUC values (0.874-0.880 at 6 months, 0.845-0.851 at 12 months).
• A web-based tool was developed for personalized risk assessment based on the model.

Conclusions:

• Incorporating HBV-specific T-cell responses into predictive frameworks enhances the accuracy of hepatitis progression prediction.
• The developed interpretable ML model aids in early risk stratification for proactive patient management.
• This approach offers a more biologically relevant and clinically meaningful tool for managing HBV infection.