BHAFT: Bayesian heredity-constrained accelerated failure time models for detecting gene-environment interactions in...

Area of Science:

• Genetics
• Biostatistics
• Computational Biology

Background:

• Understanding gene-environment (GxE) interactions is crucial for disease etiology and prognosis.
• Existing statistical methods face challenges with high-dimensional data, diverse environmental factors, and survival analysis complexities.
• The effect heredity principle aids interaction identification but lacks dedicated Bayesian survival models.

Purpose of the Study:

• To propose novel Bayesian heredity-constrained accelerated failure time (BHAFT) models.
• To incorporate the effect heredity principle into survival models for identifying main and interaction effects.
• To address limitations in detecting GxE interactions within high-dimensional survival data.

Main Methods:

• Developed BHAFT models using spike-and-slab or regularized horseshoe priors.
• Implemented Bayesian inference with the R package rstan.
• Applied models to identify GxE interactions in lung adenocarcinoma prognosis.

Main Results:

• BHAFT models outperformed existing methods in signal identification, coefficient estimation, and prognosis prediction.
• Identified biologically plausible GxE interactions relevant to lung adenocarcinoma prognosis.
• Successfully detected both main and interaction effects, crucial for GxE interaction exploration.

Conclusions:

• BHAFT models offer a powerful new framework for GxE interaction analysis in high-dimensional survival data.
• The models effectively leverage the effect heredity principle for robust interaction detection.
• This approach enhances understanding of disease etiology and prognosis by integrating genetic and environmental factors.