Bayesian Learning of Personalized Longitudinal Biomarker Trajectory.

Area of Science:

• Biostatistics
• Oncology
• Precision Medicine

Background:

• Chronic myeloid leukemia (CML) management relies on continuous biomarker monitoring for early relapse detection.
• Longitudinal biomarker measurements in CML patients exhibit significant inter-subject heterogeneity.
• Understanding these trajectories is crucial for predicting treatment resistance, but underlying mechanisms remain unclear.

Purpose of the Study:

• To develop an effective personalized prediction model for longitudinal biomarker trajectories.
• To address the challenge of heterogeneous patient data in cancer targeted therapy.
• To facilitate early prediction of disease relapse and inform clinical decision-making.

Main Methods:

• A novel Bayesian approach for modeling the distribution of subject-specific longitudinal trajectories.
• Flexible Bayesian learning to capture complex temporal patterns and non-linear covariate effects.
• Real-time prediction capabilities for both existing and new subjects.

Main Results:

• The proposed Bayesian model effectively accommodates complex and heterogeneous biomarker trajectories.
• The method allows for accurate in-sample and out-of-sample subject predictions.
• Demonstrates potential for enhancing personalized treatment management in CML.

Conclusions:

• The novel Bayesian approach provides a powerful tool for personalized prediction of biomarker trajectories.
• This methodology can significantly aid in clinical decision-making for CML patients.
• Contributes to advancing precision medicine through improved understanding and prediction of treatment response.