Bayesian personalized treatment selection strategies that integrate predictive with prognostic determinants

Area of Science:

• Bioinformatics and Computational Biology
• Genomics and Precision Medicine
• Statistical Modeling in Healthcare

Background:

• Advancements in informatics generate massive biological databases, crucial for personalized medicine.
• High-dimensional 'omics' data presents challenges for traditional statistical methods, often relying on linear models and prognostic biomarkers.
• Existing methods struggle to identify predictive biomarkers that interact with treatments and lack clarity on selection rule efficacy.

Purpose of the Study:

• To develop a robust Bayesian predictive methodology for personalized treatment selection.
• To integrate both predictive (treatment-related) and prognostic (disease-related) patient data.
• To effectively utilize complementary information from prognostic and predictive biomarkers for optimized treatment choices.

Main Methods:

• A novel Bayesian predictive method is presented to address high-dimensionality in 'omics' data.
• The methodology explicitly models the structural constraints of prognostic and predictive biomarkers.
• Case studies involving lower-grade glioma and simulations based on leukemia data are used for illustration and performance evaluation.

Main Results:

• The proposed method effectively integrates prognostic and predictive biomarker information for treatment selection.
• Theoretical analysis demonstrates the impact of prognostic features on treatment selection decisions.
• Simulations show competitive performance compared to existing methods in deriving effective selection rules.

Conclusions:

• The developed Bayesian method offers a powerful tool for personalized medicine by leveraging complex patient data.
• This approach enhances the utilization of molecular profiles for advancing disease understanding and treatment selection.
• The methodology provides a framework for robustly integrating diverse biomarker information to guide clinical decisions.